WARNING: Summaries are generated by a large language model and may be inaccurate. We suggest that you use the synopsis, short and long summaries only as a loose guide to the topics discussed.
The model may attribute to the speaker or other participants views that they do not in fact hold. It may also attribute to the speaker views expressed by other participants, or vice versa. The raw transcript (see the bottom of the page) is likely to be a more accurate representation of the seminar content, except
for errors at the level of individual words during transcription.
AI technology is rapidly changing how humans interact and think, allowing for faster conversations and insights. It has potential to expand the mental working space of individuals and touch almost every aspect of life, from extending communication across centuries to providing feedback in education. However, it can also increase dishonesty, such as compulsive lying and deep fakes, making it difficult for universities to assess students. AI is a powerful tool, but it can also be used for malicious activities.
AI technology is rapidly changing how humans interact and think, allowing for faster conversations and insights. AI tools can be used to come up with new topics for discussion, and can even replicate entire teams behind a single person. AI Shadows, which are front ends for a person's way of thinking, can be placed together to play a game and can be used to draw insights from conversations faster than real time. AI has the potential to expand the mental working space of individuals and touch almost every aspect of life.
Technology has enabled individuals to communicate with people across vast distances, both in space and in time. This is exemplified by YouTube, which has allowed an individual's voice to reach many more people than before. Additionally, AI technology can be used to augment and expand capabilities, such as by using chatbots to facilitate initial meetings with students. This potential of AI is often overlooked and should be taken advantage of to get more work done.
Technology has enabled humans to extend the reach of their thoughts and communication beyond the present moment and across centuries. The internet has allowed people to quickly share their ideas with a wide audience via media such as YouTube videos, Tick Tock clips, tweets, and more. This primitive form of replicating one's mind is not perfect, but still has the potential to profoundly extend the reach of our minds. Additionally, bots can be used to spread ideas more convincingly than books, giving mini lectures and responding to questions. This could be a new way for ideas to spread and gain influence.
The rapid advancement of artificial intelligence has enabled it to generate content at a much faster rate than humans can digest it, raising implications for areas such as advertising and books. Humans have an immense capacity for generating new content, but their capacity for digesting and understanding it is limited, making it difficult to get people to read books and other messages. This finite bandwidth of human beings as consumers of information is why advertising and promotion are necessary to disseminate ideas.
AI systems are becoming increasingly capable of ingesting and digesting material, which will reduce the need for advertising and promotion. However, attention will become a scarce resource, making it difficult for people to interact with famous people. To create a utopian world, AI agents should be used to filter out emails and discover talent, while also allowing anyone to log onto a website and have an interaction with an AI to pick up sparks of why they might be interesting. This would ensure fewer people get lost and everyone can access what they need.
AI technologies can be used to coordinate prospective employees and employers, with the potential to improve the process or make it worse. China is exploring using AI technology, which could lead to either more decentralised or centralised solutions. When a new capability or resource becomes available, it typically leads to a proliferation of niches and an expansion of diversity, meaning both decentralization and centralization become viable in the future.
Recording oneself and interacting with primitive bots can help improve communication and thinking skills. This technique can be used in education and training, such as sports, and should be customized and scaled to the ability of the student to push them to near their limit. Decentralized systems could support more socio-institutional diversity than centralized systems, but scale and the unit of analysis must be taken into account. New forms of interaction, such as intellectual dating, could also be explored.
AI has the potential to improve education by providing feedback and acting as an affordable private tutor. It could also increase social competence in those living and working remotely, and even help people make friends. However, AI tools could also increase dishonesty, such as compulsive lying and deep fakes, making it difficult to tell what is true.
AI is making it easier for people to cheat and lie, such as the industry of writing Masters and PhD theses for others. This has caused universities to rethink the education system and how to respond to dishonesty. The current pandemic has further disrupted the traditional exam and assignment system, making it difficult to incentivise and assess students. Universities must now find alternative solutions to motivate and discipline students, rather than relying on submitting assignments.
AI technology is rapidly transforming existing Human Institutions such as universities. A thought experiment is proposed in which two AI Shadows are created, one for the speaker and one for Adam. These Shadows would have access to their respective books and other written material, and would be a front end for their way of thinking. They can be placed together to play a game, and although not perfect replications of the two people, would be fairly good.
AI tools are not necessarily going to replace human agents in institutions, but instead have a new role to play. Reasoning by analogy from past experiences is often limited, as seen in Dan's example. This is because nothing in a caterpillar suggests it will become a butterfly, or an acorn suggests it will grow into an oak tree. AI tools can be used to come up with new topics for discussion in seminars, and can be integrated into the fabric of interactions between people.
Artificial intelligence is rapidly changing the way individuals work and think, allowing for faster conversations and insights. For example, two bots or shadows of existing people could have a conversation faster than real time and draw insights from it. AI can also replicate entire teams behind a single person, either in real time or accelerated time. This is an astonishing development, as AI has the potential to expand the mental working space of individuals and touch almost every aspect of life.
AI technology has the potential to augment and expand capabilities along various dimensions, such as time and multiplicity. This idea has been explored in science fiction, and is now becoming science fact. For example, academics can use YouTube and the internet to attract students. Google prioritizes YouTube in search, so by having videos online, academics can get more work done. This potential of AI is underappreciated by many people, and should be taken advantage of.
Technology has drastically expanded the reach of communication. 500 years ago, the only way to communicate was by word of mouth or by printing in small quantities. Now, technology has enabled individuals to reach many more people across vast distances, both in space and in time. For example, YouTube is an example of an older phenomenon that has allowed an individual's voice to reach many more people than before. Additionally, technology has enabled low friction interactions, such as chatbots, which can be used to facilitate initial meetings with students.
Technology has enabled humans to extend the reach of their thoughts and communication beyond the present moment and across centuries. The internet has allowed people to reach a large number of people quickly with media like YouTube videos, Tick Tock clips, tweets and more. This is a primitive form of replicating one's mind, and although it is not a perfect replication, it still has the potential to profoundly extend the reach of our minds. This is separate from the issue of generating new instances of intelligence or superintelligence, and even these fractional instances of a person's intelligence are astonishing to contemplate.
Ideas can spread more convincingly through bots than books. For example, a bot could be used to give a mini lecture on Euclid and answer questions about him. This could help more people get into topics they wouldn't normally get into. Advertising and influence can be scaled through viral videos or social media, but bots can actually engage with people and respond to their questions. This could be a new way for ideas to spread and gain influence.
Artificial intelligence has advanced rapidly in recent years, allowing us to generate content at a much greater rate than we can digest it. This is contrary to our initial expectations, as we thought the digestion capacity would have grown massively before the generative capacity. This has implications in areas such as advertising and books, as it would be possible to digest every new book quickly, but the generative capacity of AI systems will still overwhelm the digestion capacity. A super intelligent AI or cybernetically enhanced individual would be able to digest new material quickly and precisely, but this will still be insufficient to keep up with the rate of generation.
Humans have an overwhelming capacity to generate new content, and the balance between generating and digesting information is shifting. Despite this, humans are still limited in their capacity to digest and understand new content, making advertising and promotion necessary to disseminate ideas. This finite bandwidth of human beings as consumers of information is the reason why it is difficult to get people to read books and other messages.
Artificial intelligence systems will soon have vastly greater capacity to ingest and digest material than humans, which will make advertising and promotion less necessary. However, people will still need to filter and be picky about what they consume. Furthermore, with more people on Earth, attention will become a scarce resource, making it difficult for people to interact with famous people. Agents will filter out most people, but soon people will be able to interact with AI shadows of famous people, and if they are interesting enough, they will be passed on to the real person.
People are increasingly delegating intelligence to AI agents to filter out emails and discover talent. This has made the modern world more productive, but it also creates a dystopia where young people face a wall of AI's between them and anything of value. To create a utopian world, there needs to be enough AI agents that anyone can log onto a website and have an interaction with that AI to pick up sparks of why they might be interesting. This would ensure fewer people get lost and everyone can access what they need.
Job markets struggle to match prospective employees with prospective employers due to a lack of coordination. AI technologies could potentially make this problem worse or better, depending on how they are used. There are already many industries dedicated to solving the coordination problem, such as dating apps and employment services. AI could be used to improve this process, but it could also make it worse.
AI technology is making it easier for people to coordinate and extract rents. There is debate between whether this will lead to more centralised or decentralised solutions. China is looking to deploy these technologies, but it is not clear which direction this will lead. Institutions are used to coordinate people, but they are often wasteful and inefficient. AI technology could potentially improve this, but the outcome is uncertain.
When a new capability or resource becomes available to an ecosystem, it typically leads to a proliferation of niches and an expansion of diversity. This suggests that the same will happen when new technologies become available, leading to both more decentralization and centralization. Decentralization can be thought of as niches, and centralization as the power and capability to exist due to the abundance of resources. This suggests that both forms of Institution will become viable in the future.
A decentralized system could support more socio-institutional diversity than a centralized system, but scale is also a factor. The unit of analysis must be taken into account, as a small patch of land will not support much biodiversity. The planet as a whole could see more of everything, but individual societies may have more centralization or decentralization. New forms of interaction, such as intellectual dating, could be explored.
Recording oneself and putting it on YouTube can help people become better teachers and speakers. Listening to the recording allows the speaker to identify and eliminate pauses and repetitions and become more eloquent. Moreover, interacting with primitive bots could also help improve communication and thinking skills. This technique can be used in education and training, such as sports, and could be incorporated as a proactive part of education. It should be customized and scaled to the ability of the student, pushing them to near their limit, so that their limit expands.
AI can be used in education to provide feedback and act as a private tutor at an affordable cost. This could lead to everyone having a life coach, tutor, or team of AI tutors helping them reach their potential. If AI achieve a higher level of competence and social cues, they could play a powerful role in socializing instruction and providing children with social and emotional intelligence skills.
AI has the potential to increase social competence in children who otherwise suffer from social isolation due to living and working remotely. Alcohol has been a social lubricant for millennia, and may help people make friends. An AI shadow could potentially say the silly and embarrassing things that people may not be able to say themselves.
AI tools have the potential to increase dishonesty in the context of job seeking and other aspects of life. An example of this is George Santos, who is believed to have a mental illness causing compulsive lying. In the post-truth era, Donald Trump is a prime example of this, and AI tools could further fuel this dishonesty. Deep fakes and generative functions can make it difficult to tell what is true.
AI is making it easier for bad faith agents to lie and cheat. An example is the industry of people who write Masters and PhD theses for others, who are intelligent yet sociopathic. AI could make this industry even more dangerous, and it is a cause for concern. Academics are worried about GPT writing essays, which could be a small problem compared to the bigger changes coming.
AI has thrown the essay-based education system into disarray, as there is no clear response to the potential technological responses. Universities have the fallback of reverting to oral examinations, but this is not as effective as it was in the past. Other situations where dishonesty is turbocharged by AI could be catastrophic, as there is no old solution to fallback on. Therefore, AI has caused a rethink of the education system and how to respond to dishonesty.
Universities are facing a huge problem due to the disruption of the current pandemic and need to find solutions quickly. Exams and assignments are partly used to incentivise students and partly to assess their capabilities, but cheating has been so prevalent for so long that these grades are not reliable. AI solutions for submitting homework may be a possibility, but it is difficult to be sympathetic to universities that have been using a sham system for so long. They must now find solutions to motivate and discipline students and allocate their time, rather than relying on submitting assignments.
Examinations and essays are common evaluation tools used to measure student performance, however, multiple choice tests have become more popular due to the increased number of students and the need for faster grading. This is a consequence of universities scaling up too quickly, with professors and administrators taking the blame for the unsustainable growth. The bill is now coming due, with professors and other classes of human beings suffering the consequences of their misbehaviour.
I want to just assume without argument that we're on track to AGI within 10 years it may or may not be true but I don't care for present purposes let's assume that's true so we're playing the game of thinking through what's going to happen now if that's true then every Human Institution that currently exists will be transformed government educational institutions uh everything that we do every form of coordination that takes place between people will be transformed [Music] um eventually in profound and fundamental ways but initially in gradual ways and that's going to take place already this year next year we can see like you know the discussion around cheating with chat GPT in universities is like this kind of storm in a teacup I don't know it's like this very strange discussion to see people having it I mean it's good I suppose that people are waking up to these things but that issue in particular seems so minute uh to me so I want to talk about not only so last week uh well this a doctor was here and we were discussing um how these kinds of Technologies might be incorporated into seminars this week it's listening to us and reacting hopefully in a constructive way uh that's a new player right so the AI is a new contributor to an existing institution an existing discussion an existing format a seminar is a very old Human Institution uh it's embedded in something meta uni which is pretty similar to institutions that have been around for a long time what I want to talk about today at least my proposal is not just new players but new games so here's Just one thought experiment to get us started so right now we have this AI it's it's sitting in the seminar contributing to some degree here's what I can imagine happening soon so this AI is kind of like Adam's Shadow right it's it has access to his books it's kind of not only his books but some books I imagine are kind of aligned with how I think about Adam things like George Dyson's books um so he can configure that and refine it to be more like the way he thinks by referring to more of his written material and also Unwritten material and books that he chooses and so on now that will never be a perfect replication of Adam but it's kind of a front end to his way of thinking and because he's produced a lot of written text the front end may be fairly good now here's a game you could play suppose both Adam and I produce two such Shadows so I have a mini me and Adam has a mini me then we can sit those AIS together somewhere they don't have to be
represented in a virtual world but they have a conversation for six virtual weeks and then they distill all that down to the four topics that will be most productive for Adam and I to discuss in the next seminar what if you come to every meeting with another person having suggestions for the Deep areas of alignment that maybe you're not aware of that would be profitable to discuss which you have not discussed in the previous 60 hours of disruption seminars that's totally doable we could do it right now that's an example of a new game right that's like a kind of interaction between people that just does not exist currently so I think we can maybe talk about that as an example but more generally I want to discuss what's going to evolve as these Technologies become more and more deeply integrated uh not only as tools but into the very fabric of interactions between people do we want Shadows like this yeah that I think that this is this raises an interesting point about the limitations of what reasoning by analogy and or uh reasoning via simple extrapolation from the past it this example of yours Dan reveals some of the limits of that and we do see similar limits for disruptions of other kinds where the uh uh a a nice way of um a nice metaphor for that is that's been evoked in the past by others is uh I forget the exact quote but it's something like nothing superficially nothing about a caterpillar uh suggests that it's going to become a butterfly and other metaphors have been invoked as well you know the the the the acorn doesn't on its surface look like anything like the oak tree that it could that it will grow into and um so that there's just a there's a this is an old standing sentiment a long-standing sentiment that um uh past experience is not necessarily a good predictor of the future and that reasoning by analogy from its past experiences is often very limited and I think that you've you've so anyway with that framing in in terms of general principles I think the specific example and I'll draw another one here as well to go along with it but the specific example you you noted Dan is uh where these AI tools are not necessarily going to insert completely newer independent agents and are not necessarily going to uh replace on a or substitute on a one-to-one basis human agents that are that are in institutions already but rather they have this entirely new role to play uh and this is one example of something like that you described maybe I can think of a or we can think of some
others but in the role that you described this is this new augmenting role has has you know if it has any of the features of of standard agents that we're familiar with it you know it doesn't it doesn't share all of those features that is something that some things that are very different um so uh it's it again it's quite alien that uh you could imagine these shadows as you're calling them these sort of partial agents or unconscious um uh intelligence intelligences uh them entering into a function in which they they effectively telescope the the the time scale upon which a person's um mind is able to work and draw insights and and and and uh you know generate um generate content from which to stimulate the production of knowledge or other action and in this case you talked about you know uh you could you could take two Bots Shadows of existing people and have them have a conversation faster than in real time and draw insights out of that it's fascinating thought experiment you know a marvelous possibility to imagine and probably something that's going to happen very soon and be extraordinarily impactful another example of that uh kind of pivoting that maybe um uh by 90 degrees from time into space is you know you could you could imagine that spatially too why be why have it be limited to Just One agent why not have you know uh two or five or ten or Twenty or a hundred agents having that conversation and in other words why not replicate entire teams behind you in and then and uh that could be either in real time or in you know accelerated uh uh time and and so there's there is what we see is just an explosion of the uh the what you call it a working space maybe not a possibility space but perhaps an explosion of the of the mental working space around individual people individual human beings today and these are these are not General AIS you know it's it's it's the potential of narrow artificial intelligence as we've defined it in the past to to explode those boundaries of what uh just through augmenting with these extraordinary intelligence Services augmenting what an existing person could do or could achieve um and it is an astonishing thing to think about uh and there are so you know there are implications there's almost an end almost an endless supply of implications to think about because the real challenge is trying to imagine anything this does not affect in our lives I think that is the harder challenge can you think of anything that will be that will remain Untouched by uh the the
disruptive um impact of of the of this new artificial intelligence technology um so I don't know if there's a whole lot of substance there but the uh I think we could we could maybe we could then try to wrap this in a in a framework with with maybe some categories like what are sort of dimensions of performance that you could augment if you could if you had the capacity to um multiply certain intellectual tasks as instances of your own faculties or or partial instances of your own faculties you know along what dimensions could could you expand or explode that uh capability Dan you mentioned time um I think we can do that in terms of you know space is maybe not the right way to think about it but certainly just multiple multiplicity multiple instances instead of just one of you I mean that's that's that's you know and this is a staple of Science Fiction right uh even even before even not necessarily with with um these kind of AI in mind if you go back you know Generations in science fiction you get ideas of like people cloning themselves so that you know there are two or three of the same person and they can get more done in the same amount of time I mean these are these are fun old ideas that have been explored somewhat in this in the science fiction uh Arena and now here we are you know as ever science fiction becoming science fact um let me draw a line through a few dots to do with this there's an idea so this is already true today of uh say YouTube and the internet uh so I'm despite these Technologies like YouTube and web pages being at this point ancient uh very few academics take use of them to a great extent uh in the sense that you know the best way to attract students is to have videos of the you do on YouTube right right because right Google really loves YouTube it prioritizes it in search uh you know I'm not spending a lot of time uh going out there to look for students students come to me and I don't know exactly how they find me uh obviously you know the undergraduates at the University of Melbourne they interact with me in a class they find me um but you know people get in touch with me from around Australia and internationally they've come across my work usually it's through Googling various terms or through you know hitting on one of my YouTube videos or something so the work you can get done by just having the stuff you do out there for the current generation of AIS to look at meaning like old school AIS things like Google search uh is underappreciated by by many people and
that's a form of parallelism right so there's there's like a pretty weak non-interactive version of me out there in videos talking a lot of the time and through that it's kind of like a filter that directs people to the real me uh and of course I have very little attention compared to that those scalable versions of me so you know I can't take a thousand students and nor do I want to but um but in terms of uh yeah I mean having parallelizable versions of you in the form of a chatbot is a kind of extension of that idea right but a much more engaging potentially useful form which gets close to being able to do some of the initial interactions so usually what happens when I meet a student for the first time is that I say a bit about what I do I go through many of the things I do and I kind of see what they're interested in and I know a bit about their background so I focus on the things I think they might be interested in that could pretty easily be done maybe already by the systems that exist or near-term systems I'm not saying I want to offload that to the AI because it's my initial meeting with a student right so it's it's part of the process and I like it and I enjoy it but uh it doesn't seem to be harmful for that to be available in a very low friction way just sitting out there right so that's the kind of thing that will be you know I could do that right now yeah that's that's a yeah it's amazing to think through that let me back let me let me take a step back to something you said earlier um you mentioned YouTube and YouTube is an interesting example of an older phenomenon which is the the uh radical expansion of reach that technology gives to communication to voice to an individual's voice an individual's you know the information emanating from an individual 500 years ago um the you know the with with without telecommunications the bandwidth uh with which with which any individual person could communicate to others was severely limited you would have to pass something word by mouth um or Word of Mouth uh for the most part uh and to a far lesser extent 500 years ago it was just becoming possible to print in large quantities which of course extended the reach of communications not just in space and in in you know across vast distances but also uh numbers of instances so you could pass on By Word of Mouth a message you know through uh many other speakers and potentially through copies of your book and then also it extended reach through time so a uh thought that a person might have
written down on a book could potentially reach across spans of centuries to to others and and communicate in a way that that uh just vastly vastly exceeds what uh human beings were were evolved to be to have the capacity to do which is simply communicate with voice one time in the present moment uh so I'm reminded of that when you said YouTube and and you mentioned professors because of course it's possible to reach huge numbers of people people very quickly now through the internet and with media of various kinds there's a YouTube video or you know a 15 15 second Tick Tock clip or a tweet or whatever it might be but those are instances where you are where we have in the past we and we've all lived through the through that well not we all but but many of us alive today have lived to see the the the internet arise since it's in Inception um uh in the early 90s for practical purposes from the public perspective and now we have I think an analogous uh situation where the it isn't reach but it's it's something similar of instancing of your thoughts being having the potential to be similarly expanded across time and space and instance I don't know how to describe how you would describe numbers of instances but but instead of a particular communicating a particular message in this case it's the reach of your your uh cognition your cognitive you know your mindset your cognitive faculties and it and these are primitive Technologies so as you said Dan it's not a perfect replication of one's mind presumably we will get there at some point uh where you have where a person might have multiple instances of their entire mind um uh existing simultaneously but this is even though this is just a shadow of that it does it does it does point to a a similar kind of profound extension of the reach of our minds in the same way that there was a profound extension of the reach of our thoughts or our Communications our speech in um I guess as it were in the past and this strikes me as as well I I I I I can't think of anything more profound um a sort of lack vocabulary words to describe how profound I think this this increase of reach has the potential to be and this is all separate from the issue of of generating fundamentally new instances of intelligence and uh uh let alone generating instances of super intelligence so this is all just generating instances of sub-human or partial human fractional instances of a person's intelligence and even still it's a it's astonishing to contemplate
the implications of that yeah I think it's interesting to think about why ideas win so you were talking about the ability to scale ideas or have them survive across time which is a kind of scaling right so your idea gets into the minds of people year after year after year if you have a book that is printed many times and widely read um but a book is sort of requires a level I mean if it's a complex idea it sort of requires something from the reader right and if they get stuck then you lost them you could imagine a future in which uh the ideas that influence Society are the ones that are packaged most convincingly into Bots right not only I mean that's very close of course to just like the Bots that exist on Twitter right uh the the kind of that are just spamming the new sphere with your ideas um but there's a version of that which uh I don't know if it's uh on the light side exactly but um just to take an example so uh fairly soon I'm going to augment the version of me that stands at the entrance to meta uni with a bot capability so it will it will give the little mini lecture on Euclid and then it'll be pretty tightly constrained to only talk about Euclid and Maddie any stuff and it'll have a kind of script that it can query from but it will answer questions about Euclid and explain why you might care about Euclid now there aren't a lot of people walking around trying to convince you to give a about Euclid right so uh having that bot standing there for every random Roblox person who walks past advertising Euclid um well first of all that's a kind of model for what advertising will look like uh but it's it's also like if you want your ideas to spread it's probably not enough to write books it's already not enough to write books right books are relatively yeah they lost their influence but um I mean I don't I don't know how it it's at current prices it certainly can't compare to the scale of influence you get from uh you know viral videos or you know Facebook or social media in general but see I'm but it's different right because it can actually engage with you and uh respond to your questions I can imagine many more people getting into topics they wouldn't otherwise get into because they happen to run into something like this that's walking around foreign yeah this makes me think of another this is this is a tangent forgive me it's it's not perhaps directly related to the to the discussion here uh it seems to me perhaps it's a little bit of an irony maybe a little bit of a bitter irony and
one that I would that surprised me um and that is that the our capacity to we use these new technologies our capacity to generate content vastly exceeds our capacity for now to digest content and ah naively if you if I project in my mind back to my earlier self from 10 or 15 or 20 years ago I don't think I would have I think I would have been surprised at again the generative capabilities of artificial intelligence that we've seen I mean The Marvelous astonishing ones we've seen just in the last several years uh you know subsequent to the transform Transformer and um uh diffusion you know approaches to these challenges success is simply astonishing but if I if you'd asked me 10 years ago I would have I would have um thought it would be the other way around that that the ability to read to digest um a material would have uh outstripped our capacity to generate it in you know or if there was a sequence that the digestion capacity would have grown massively before the generative capacity that would have simply been my initial uh naive intuition and it turns out I think at least as far as I can tell that that is that that seems to to not be the case seems to be the other way around um but the bitter the thing that's bitter about that in this context and I'm thinking you know you mentioned advertising Dan in your example of a bot you know standing on the corner of or being somewhere um where people can encounter it and you mentioned also the example of books it's really unfortunate that that I mean one could imagine the capability of digesting every book every new book that's written it just you read it because it gets digested um if if you have the you know the the capability either um at some remove from your own mind or perhaps increasingly integrated directly with your own mind or in the case of a fully artificial mind an artificially sentient general intelligence should presumably be able to digest you know new material fairly straightforwardly at a scale and with a speed and with a Precision that that certainly superhuman god-like as far as human beings relative to human beings are concerned but unfortunately what it looks like is that the generative capacity of these systems will still overwhelm the digestion capacity in other words if if we were only producing books at the speed at the rate at which civilization was doing so in the year 2010 then an artificial general intelligence or a cybernetically enhanced individual uh who was super intelligent might perhaps aspire to
digest you know and read and understand and and and integrate into their mind uh every new book written but now now that you can press a button and have a new book you know uh generated instantly and that will only become easier uh the the the the the it's it's it's I'm still grappling with the shock of the fact that um we we you know despite Godlike uh capacity potentially to digest uh information that the there is nevertheless going to be an overwhelming Force capable of generating new content and um I I don't I mean I don't know if the tables will will turn or ever turn or we'll turn anytime soon or anything and flip in those the balance of those two things flip but I I don't know if if you're struck the same way that the the gender of the generative capacity being so vastly um uh uh powerful so soon and I don't mean just just nonsense I mean I I'm not talking about just just nonsense books being published that have that have no content in them I mean we could do that just with noise you could just publish you know an infinite number of books full of random characters that's not what I'm talking about I'm actually talking about proper content in the same way that that you know you could you could now cons for all practical purposes produce an unlimited unbounded shall we say um amount of new art and people's people may nay say that and say well it's you know this it's it's it's nothing fundamentally new about it because it's it's derivative but I think that's a bit nonsense and in in some sense everything has to be derivative um everything is derivative plus a little bit of noise from the environment as far as I can see um and uh so I just find that shocking I I'm I'm grappling with how unintuitive I still find that uh um this situation where we're absolutely swamped by a new content and it far far outs are outstrips our capacity to ingest it and digest it and the reason why I think this is relevant just very quickly to wrap to Circle back and wrap up this thought with respect to advertising and and with respect to the dissemination of ideas say trying to write a book and then getting people to read it which is very difficult these days um the the the only reason why that is difficult the only reason why advertising is necessary or as a function and and the only reason why um people put efforts into promoting books and any other message is because of the finite bandwidth the finite capacity of of human beings as consumers of information and other content they're
very very finite very very limited capacity to ingest and digest that content now if you had a massively greater capacity you know five ten orders of magnitude greater capacity to ingest and digest material as it seems some of these artificial intelligence systems will have then there would be no point in advertising anything there would be no you know there would be no scramble to to promote your book to try to get the message out because as soon as you wrote it everybody with that capacity would read it because everybody would read everything and that I was I guess naively um based on the science fiction leading up perhaps to the year 2000 or so I was imagining maybe we would get to that kind of future now it sounds like unfortunately the the the advertising and the promotion is going to continue to have a role because there will just be that much more new content that we will still have to filter and still be picky and choosy about despite vastly increased ingestion and digestion capacity so not that related to the conversation I apologize for the for the tangent but fun idea anyway yeah I think if we take if we take that a little bit sideways uh here's what I think will happen fairly soon uh with regards to not filtering out books but people so if you think about now we take for granted the kind of we have this problem with this seven billion people on Earth uh there are limited resources in terms of interesting well-paying jobs or even just attention which is going to become the scarce thing right the the attention of people with interesting ideas I mean how many people would love to have a conversation with Stephen Pinker well it was probably Millions right Stephen Pinker does not have that kind of scale uh and well for various reasons you don't just want to interact with his ideas right it's uh the people who want to talk to Stephen Pinker don't want to only interact with the Stephen pinkerpot right of this kind of status or some kind of like personal uh interest in in having that interaction and making an impact on the real Stephen Pinker um now right now how do you how does that work I guess famous people have agents you reach them through their agent they filter out 99.99999 of all the people who want to talk to you uh you know maybe you you meet them at some conference or something but quite soon what will happen is that you'll just interact with their their AI Shadow right and if you're sufficiently interesting you'll get passed on to the real thing
so people will start to surround themselves with layers of kind of delegated intelligence which are increasingly good and I mean the probability that a shadow of me could tell with a short interaction whether or not I would be truly interested in talking to someone is very high right it's not that difficult to figure it out I mean I get email every day from somebody wanting something that I'm just not interested in talking to you know every academic gets spam emails for lower quality conferences they should go to or there are students in developing countries no offense to them but who are just Mass emailing every Professor on Earth to look for a PhD position right that's I mean if I mean I have friends in those kinds of countries who have done that and I understand why they do it it's not effective certainly yeah pursue some other approach but uh still yeah I mean I get Amazon like that and I I do reply to all of them and say sorry I think most people don't even reply to them okay so that's that's how email currently works and we we have all sorts of layers of spam detection and other filtering to to restrict that onslaught of email and I'm not a famous person I'm famous people must get thousands of emails a day that they have to filter through and they have agents to do it um but it would be I mean I'm it's a fairly dystopian I think but maybe it's just like that's a a status quo bias as well right you could say the present is fairly dystopian where there's just even if you're I mean there are many brilliant young people out there who just cannot get in touch with anybody who would understand them and who would make use of their gifts right uh there are people uh out there trying to look for such talent and we have various mechanisms in place to to filter that Talent up and make use of it and that's a large reason why we're much more productive in the modern world than we used to be that we actually discover that talent and properly well to some degree properly allocated but okay so the utopian part is what if there's just enough agents sitting around that anybody anyway can log onto a website and have an interaction with that Ai and it picks up some spark of why they might be interesting and and then so you there would be much much fewer people who get lost on the other hand that means that as a young person you're just facing this wall of AIS that are in between you and everything of value right you will not get a position unless you convince five layers of ai's
associated to various institutions it's like a never-ending test uh so and I I should I should point I mean the truly you already mentioned PhD uh individuals pursuing PhD opportunities and the nest the necessity of effectively spamming or you know pick your metaphor but but um just sort of shotgun approaching approaches or carpet bombing approaches to just just just sending out huge numbers and hoping that one or two land or stick or something like that the current job market is very much like that where there's a there is to some extent a massive coordination shortfall a massive coordination dysfunction or problem um certainly globally and I I presume it's it's quite terrible even within societies where we really do struggle to match uh uh prospective employees with prospective employers and uh that is in the in the very real world of of just the realities of the of the 21st century job market for for labor seeking employment and the of course we can we can broaden that into a larger perhaps more abstract category of anyone seeking an opportunity struggles to be matched with anyone who has an opportunity to share and is looking for talent of some sort to seize that opportunity and that could be you know students looking for uh opportunities employer employees looking for opportunities institutions looking for opportunities you know this is the the our inability to coordinate within the the larger opportunity space of our civilization even if that we've made vast improvements over where we were a generation or two ago perhaps we're still vastly sub-optimal and a key an interesting question in the abstract I think here is uh well two questions one descriptive and one normative the descriptive question it would be will these new AI Technologies make that problem exacerbate the existing problems uh make that coordination challenge worse or or will they make it better and then the normative question of course that follows is okay how might these new technologies be used to solve some are all of that coordination problem and and we obviously have very large Industries dedicated to and making great grandiose claims about being able to solve some portion of the coordination problem uh between opportunity Seekers and opportunity providers so dating apps are a great example um Employment Services agencies are an example there are many others right and one could imagine artificial intelligence serving both functions serving to make this entire problem vastly worse and also at the same time perhaps really
making a big difference in in in superhuman capacity to achieve coordination I don't know which side is more plausible certainly cynics would be inclined towards the pessimistic pessimistic outcome there I suppose you can certainly see cases being made there but anyways it does strike me as something quite fundamental here and perhaps a very early application because presumably there's a lot of money there's a lot of opportunity to extract rents on either side of that coordination problem in other words extract rents from people who are seeking opportunities and extract rents from people who you know people individuals agents agencies institutions whatever level of organization but extracting rents from people who are providing opportunities as well and I know that I've just been involved in in the um as as a potential employer on that on one side of that it's very difficult to find the talent that we were looking for it was a serious challenge um so yeah what do you think's more likely centralization or decentralization uh when it comes to the redesign of Institutions so imagine people start taking on both these Technologies the the pressure I mean there's pressures in both directions right so uh there will be new cost Associated new costs associated to baking AI into everything uh pervasively those costs will be easier to pay for centralized institutions um on the other hand I mean when you read I mean there's a long old literature about uh negative takes on centralized education systems whether it's primary high school or university that Envision a more anarchic system in the positive reading of that word um where people come together and teach each other without a centralized Authority kind of extracting rents from that process you could imagine that being super powered by these kinds of Technologies in various ways um and more abstractly the question is about coordination so you're referring to it earlier I mean that's what we use institutions to do right how do we coordinate thousands and thousands of people to push in One Direction uh consistently and over long periods of time we're really bad at that somehow so we have Solutions they're better than Solutions in the past but they're still extremely wasteful uh it's not clear to me whether the new forms of coordination that will develop uh naturally more centralized or less centralized they can see Arguments for both I mean you can look at the way China is looking to deploy these Technologies and see how it could
clearly enable a more centralized form of coordination but on the other hand putting more power in the hands of individuals and allowing them to sort and find each other much much more efficiently seems like it could also lend itself to a more decentralized form of Institution I hope this is not a cop-out and is actually a legitimate response to that question but I and there's danger and Reason by reasoning by analogy here as I mentioned earlier but I'm going to Reason by analogy here um and say that the analogy I would look to here are ecosystems and typically what we see when a fundamentally new capability or new resource more broadly speaking becomes available to to an ecosystem like if we if if this ecosystem suddenly has more energy in it or um suddenly has an influx of a new uh um a a new nutrient for example what we tend to see uh in in very large ecosystems is not a a not necessarily a uh singular convergence towards one or the other a a form of life but a proliferation of niches and just a general expansion of diversity and so so when we increase the the energy in an ecosystem the energy available in an ecosystem for example what we tend to see is more biodiversity in that ecosystem eventually that there are there are exceptions you can have ecosystems collapse into less diverse structures with fewer species and um once in a while you can have an individual species just run run be so successful that it runs roughshod over the entire uh system ecosystem and uh destroys niches but this is all a way of saying that that what I suspect will happen and it sounds like a cop-out answer is that we will have both more decent more decentralization and more centralization and so how how could you how could that happen how would you have that isn't that a contradiction well it's not a contradiction if you think of it like an ecosystem if you think of of decentralization as uh uh as as niches then I suspect we're likely to see more decentralized institutions emerge and be viable in other words be able to survive in the future because of these Technologies and other quote unquote resources namely other forms other other supporting abundance and prosperity making it possible to Simply exist and at the same time at the same time that we see that proliferation of of diversity of Institution and that I think broadly aligns with the idea of decentralizing um and anarchic um uh structure I think we will also see an increase in the power and and capability simply because there's more resource and
available uh two centralized authorities and so I I suspect we're actually going to see both strangers that may seem and there and and uh because the system will be able to support more and the analogy here would be more biodiversity but in this case it would be something more like Morris socio-institutional diversity within the socio-institutional system as opposed to ecological diver bio ecological or biodiversity in ecological and biological systems I hope that makes sense um and again I could be wrong there are great dangers in reasoning by analogy across systems that are that are that are not comparable but uh that would be my that's my suspicion I quite strongly suspect that that's going to be the case and I so my money I would bet against anybody who's putting all their chips on on one or the other now I should add another caveat to that which is that it depends on your unit of analysis and I'm taking the entire planet as the unit of analysis if you're talking about an individual Society then I mean and again scale is relevant to biodiversity if you take it if you're talking about too small a patch of land area like a square meter there's only so much biodiversity you're going to have there and I think it's the same for for I think scale uh it applies similarly to these social institutional systems as well and so you've got to have Geographic scale and everything that that entails physical scale scale of inputs scale of energy scale of materials scale of number of Agents human and otherwise just sheer scale across all the dimensions we can think of you need a sufficient scale to support any kind of uh diversity at all um so it could be that within individual societies you know there will be massive centralization in one Society maybe more proliferation or decentralization of institutions in another Society but as the planet as a whole I expect to see more of everything mm-hmm yeah that's interesting I'm gonna have to go a little bit earlier today so maybe we have a few minutes to wrap up uh sorry for that um that's okay me too yeah so any other thoughts on new games I mean I guess what we talked about perhaps were new forms of interaction uh intellectual dating where you send out delegates to figure out whether you really want to spend an hour of your precious time to talk to the real thing I mean I wonder maybe I have one idea go ahead Dan I have but I do have one idea in answer to that new to the alternate games but go ahead please I feel like this will make me
a better thinker for the following reason recording myself and putting myself on YouTube made me a much better teacher and speaker first of all because I got to actually experience the lecture from the other side I could listen to it I could see how many stupid pauses and repetitions there were and I made an effort to eliminate them and I'm much more eloquent uh interacting with people in this kind of format than I used to be as a result of that interaction with a simulacrum of myself namely the YouTube video which is a very primitive simulacrum I can imagine that if I produced something like one of these Bots as primitive as they are right now uh interacting with it might make me a better Communicator a better thinker and a better speaker there's a bit of like competitive pressure I guess eventually I might lose but until I do it'll keep me on my toes I think that has the potential to play a significant role in certain parts of our education system certainly it's one of the roles that Bots could perform at various stages throughout a person's education Primary School Secondary School higher education that the your your receiving a form of feedback about your performance that could be very useful holding up the mirror the proverbial mirror as the a YouTube video of yourself does and and uh a record seeing a recording of yourself can very definitely do uh education training of many different kinds makes use of this technique uh of recording yourself analyzing it and then practicing Sports and Athletics come to mind the the idea of individuals and teams reviewing their own performance that's been recorded and um is quite a prominent feature of professional sports for example um so I I could see that playing a role in people who are people who have initiative individuals who have initiative like you do of course damn we'll have already been availing themselves of that uh that capability just with primitive like you said with premium the Primitive simulacrum of recordings of the past but if we were clever about it if we were smart about it we would we would make that a proactive part of Education going forward and increasingly interactive and challenging and provocative and scaled to the ability of the you know customized and scaled to the ability of the person uh of the student I'm thinking so that they can be pushed to the appropriate degree like however whatever the optimum is for an individual person to be challenged to near the limit of their ability so that that limit expands we can one can
imagine using these tools very effectively to do that at least part of that will be showing people where they currently are and what their current limitations and and weaknesses are so that they can improve sorry I missed a little bit of that conversation sir I risk um saying something that's already been said but um but if you're talking about uses of this stuff in education then I don't necessarily see how making a shadow is going to be like the killer um kind of feature that from an educational perspective but um just the ability to have sort of someone um directing your um sort of learning and providing it with examples and feedback um because having an AI is is going to allow um you to have basically like a private tutor at a much more affordable cost than a human private tutor so um that seems like the killer feature because feedback is kind of King when it comes to education and not just school education but sort of learning anything um and so my contribution I think um sort of brainstorming ways that the world might change would be that in an Ideal World you know I think it would be a positive change if um what came out of this was um everyone having like a life coach slash tutor slash maybe a team of of these AI tutors um that's sort of helping everyone develop to the sort of best of their potential um I don't maybe maybe sort of talking with a simulation of yourself could be a part of that but I think just the the capacity to give attention to everyone's development and like put them in their zone of proximal development like you mentioned Adam um that seems like the killer feature that would be enabled by this yeah I I think I agree let me add one element to that which is that if these if these AIS were to achieve a higher level of competence than we see today and we're approaching general intelligence uh or something convincing more convincingly close to that such that they were socially competent and could provide valuable social cues then I can see a very very powerful role for learning buddies that are AI who also play a very prominent role in socializing instruction and uh because children children who haven't who have are who are naturally pro-social and have uh social prowess things that have been categorized as social and emotional intelligence for example those are skills some people struggle to acquire those into adulthood some acquire them more quickly than others some children are blessed with having very strong skills from a very early age
but in some sense the individuals who are successful with socialization and acquiring those social skills I think I think to a large degree they do benefit from having some sort of internal model monologue or not monologue dialogue or modeling where there's a there's a little bit of a conscience it's kind of what we mean when we say well what's your conscience telling you you know what is the little bird on your shoulder what is the angel on your shoulder telling you uh that you ought to do what would be the thoughtful thing to do here what would be the considerate thing to do oh maybe you ought to feel maybe you ought to think twice before saying something mean like mean like that next time well how would you feel if somebody said that to you well if you had one of a socially competent uh learning buddy along with you that could at least have the potential to increase the social competence of children who otherwise might suffer from the potential of increased social isolation that goes along with you know Tech these new technologies like living and working and schooling remotely being entertained remotely um you know not going out under the playground and interacting with other kids in person but doing it from behind a computer screen through you know twitch and Discord and so forth you know it's easy to become unsocialized and and socially unhealthy especially as a child and lack uh social appropriate or ideal Social Development and so I can see a a pretty powerful role for AI there as well potentially and then I did have one other suggestion sorry but but I'm going to go ahead and comment down and then I do have something quite different I was a silly comment but uh don't you know the best way to make friends is to drink Coca-Cola as your AI buddy will tell you well the funny thing is that is that that's just a silly advertising line um but uh but alcohol is different I think there's some legitimacy to that alcohol as a disinhibitor has been a social lubricant for for Millennia right and and so I think it I think drinking a beer might actually work wait wait wait wait I thought you were going to say that you'll have your AI Shadow with you not to Faithfully represent your thought but to say the silly things the embarrassing things that'll endear you to other people but you can't bring yourself to say because you're you're not drinking so for people who don't drink maybe having a drunk bot it's your unfiltered self no okay so here's something funny maybe we'll all
be like uh swooning kings and queens and princess and princesses with courts courts of people courts of attendees flanking around us and of course one is the gesture right the gesture has permission to say things uh that would otherwise be you know taboo that was one of the roles of one of the roles of that social institution and the person in that role right I hate to perhaps close if you Dan you have to go um on a pessimistic note but one thing that I am very concerned about and it goes along with advertising it goes along with job seeking uh and job um uh advertising uh opportunities of all kinds people people looking to to solve those problems um and then of course in so many other aspects of life we face this problem of dishonesty of lying and and cheating through dishonesty cheating is a different you know there are many different ways you can cheat but but one of them one of cheating in the context of trying to seize opportunities in a competitive space is misrepresenting the facts presenting a a partial version or a Rosy version or a patently false version of yourself and trying to seize opportunities and win the game and you talk are asking about how games might change that way and we have a truly spectacular example of this in American politics right now with this with this individual George Santos who's that's not even his real name and he is a you know the media coverage of him is hesitating to say it but he's very clearly uh has a severe persistent mental illness that involves um compulsive lying and people are just saying well he's a con man and that kind of thing but it's it's it's I think I think it's probably quite a bit worse than that this is a person who is pathologically incapable of of uh of being honest or incapable of of not lying and um uh so anyway this is why this is a little bit fresh in my mind as we're seeing we're we're faced with this issue but of course social media and and um the so-called post-truth era that we're now living in Donald Trump being a magnificent example uh is this is very concerning because AI tools could you know just just be dumping gasoline on the house fire of dishonesty that already exists right and um uh the the the potential of AI tools to help an individual be dishonest to help them lie to help them cheat through lying and dishonesty seems to me quite staggering and there is talk about what about deep fakes and what about you know um you know generative functions that you can't tell what the truth is well that's a separate
problem and a serious one but I'm speaking specifically about the ability of AI to help you lie and uh uh bad faith actors bad faith agents who are willing to be dishonest and in some cases perhaps just out of desperation um are going to make use of these tools and the two these tools will really really help well they were really up The Lying Game right I mean if you're trying to lie on your resume or lie in a cover letter or you know fake it till you make it in some way or just lie to a client or you know just be dishonest in any way uh chat GPT is going to be a great asset and um I'll give you one specific example uh Dan I don't know if you've ever encountered this possibly it's rare in mathematics but there is an industry where you can hire people to write your coursework and in in particular write your Masters and even PhD theses they're dissertations you can pay somebody to write your dissertation for you and um I've seen you know sort of AMA ask me anything sessions among the people who work in this industry and you know they're how they justify being so dishonest and helping people cheat and it's it's quite fascinating most of the people who work in the industry are very intelligent but sociopathic or Psychopathic you know so otherwise you wouldn't be a criminal basically like that or help people other people who are criminal um or cheating you know in a massive way and they get paid a lot of money and one person was you know so I very much as a game entirely as a game and thought it was hilarious that he could spend you know one working week cranking out a totally PhD thesis a dissertation for somebody in a subject he didn't he knew nothing about beforehand and have it be accepted and the person graduate and have it all be a sham and a fiction but one could imagine I mean if there's already a human industry with human agents willing to act in that capacity imagine what that will happen when AI can do that so again kind of a little bit of a depressing note but it's I'm it's a serious cause for concern right yeah just sort of started this conversation with Dan's comment about how academics being worried about gbt and writing essays I'm not sure Dan if you were saying that that's like not a problem or if it's just like a small fish compared to the big changes that are coming but I mean it's a calamitous problem for I mean I'm potentially I'm applying to be a tutor in a subject at Melbourne uni uh in this coming semester where part of the assessment will be
essays and like this is that is like there's there's no response there's no clear response yet um to me it feels a little bit like open AI is just thrown the entire essay-based education system in the deep end um with no defense against this I mean there's like potential technological responses that you can do but I don't know they're like so in development it's not clear basically we just have to rethink everything with like a few months notice and that's that's really tough this is a big thing yeah it is interesting although sorry Dan just one quick thought um in this instance I do think with in the case of the universities in particular and education in general we do have a fallback and uh a solution and that fullback solution is to just an earlier time and era of doing things differently in the past you were examined orally you sat down I mean the tri-post exams are still conducted at the Oxford universities you sit on a two-legged chair and you are intensely grilled by your supervisors by your professors and uh that is how you pass or fail at least for parts of the program that's not to say that essays have not been part of the educational process for a long time but I I suspect we are going I mean what is the alternative I think we're going to have to fall back to interviews and oral examinations um which will be you know for for a while difficult to cheat on and fake now once you have a neural implant that's you know putting the words in your mouth then you know uh that doesn't work either but then I suppose at that point who cares I'm you know if you can do the work you can do the work so um and you could at least partly make that argument even now but at any rate that we do have this we do have this this this previous this older solution we could revert to you sub-optimal as it is in some ways um but for but for dealing with lying like like you know okay the president of the United States just gets up on stage and just just nothing coming out of his mouth is true and all of that is being facilitated by you know AI that are cooking up convincing nonsense this is a terrifying situation um right and and uh how would we respond to that well there isn't the same kind of old solution to fall back on and that's in in that another I mean I just picked that one out of a hat but other uh situations where dishonesty um turbocharged by AI would be you know calamitous so in the case of higher education I don't mean to be flippant and say oh it's no big deal I think it's
a huge problem gigantic problem and it's and and you're right that a major part of the challenge is that we are going we are it's been so disruptive this isn't something that's slowly creeping up and slowly rolling out over many years it's you know from one semester to the next uh we're gonna have to solve this problem in any any University that doesn't is is in real trouble I mean my take on that is that uh essays and a lot of the assessment we do I mean you have to separate uh actual assessment where the your trying to tell whether someone has mastered something versus an incentive system right the examinations and assignments they're partly just you know a means of getting students to pay attention and allocate time correctly and partly they're about actually assessing people's capabilities but the assessment part has been cheated on by so many students for so long that I just I don't believe I mean the the grades students get on assignments I don't believe them at all the distributions are so different from the exam distributions that so many students are cheating it's sad to say uh that I just don't care I I don't know I don't think they have any value or very little value they have some value in terms of incentives and that will be sad if that is disabled because people just paste it into chat GPT and get out an answer um so maybe we need to do a better job of finding real solutions to motivation and discipline and helping people to allocate their time rather than this crutch of getting people to hand in uh stuff every now and then but yeah I agree it's very short term a very small amount of time to be able to find the solutions to that but I I find it a little hard to feel empathetic for a lot of the universities where it's always been a sham and now the Sham is uncovered to some degree but like you knew it was a sham you've known it's a sham for 15 years I'm not saying the particular subject you're teaching is a sham but it's like uh yeah I it's um and also I find it hard to get excited about it when there's so many more things coming very soon um of course I'm teaching calculus one next semester so you know I maybe would have a different tune if uh if the AI could be submitting their homework but they were already solving their homework on Wolfram Alpha and just handing it in so I'm like what's that well and and part of it is the the Dan you mentioned not feeling too sorry for the universities um and and like to some extent this is they've made their bed and they have to lie in it
uh the I think the the this idea I agree um sorry let me back up examinations written examinations and essays are tool evaluation tools that scale pretty well it's and especially the multiple choice ones which are very common in the United States essays have fallen out of fashion and multiple choice tests uh are the norm long answer and Essay answer even short written answer questions on exams are are um they still exist but the larger the class and the greater the scale that's being attempted the more likely you are to have a an examination format that is um dehumanized and depersonalized and the multiple choice question format is sort of the that in the extremist you know that's that's it in extremists uh and of course those tests can be machine graded they don't even need to be graded by a human being uh right so um at least part of this problem is a consequence of the University's scaling and growing too big for their britches and taking on more students then they can properly instruct and evaluate now if you if your student to instructor ratio was one to five then you know that would be a very different situation you wouldn't need multiple choice tests and one instructor could grade all of their five students work and on a very personalized level and evaluate them very closely and use a variety of different tools but when it's one instructor and 500 students there's no way to grade 500 essays in any reasonable amount of time so of course you have to use multiple choice tests and that is not the fault of the professor and to some extent the students are victims of that scale I think they we can put some blame with the professors it's very convenient for academics to say that all the problems are due to the administrators but uh our salaries have been going up very rapidly there's many more professors than they used to be now the increase in the number of Administrators is even more dramatic but you know there are a lot of well very well paid professional jobs out there being professors built on the back of scaling up universities to a degree That Never Was sustainable and now the bill is coming due and I think professors as a class they can you know we can take plenty of the blame for that uh we were happy to go along with it uh so I think that well as will happen to every class of human beings over the next 10 years the bills are coming due for all our misbehaviors and uh and we will line up one by one to collect that bill and uh cry about it okay on that note uh on that note okay