AI AGENCY ISN'T HERE YET... (Dr. Philip Ball)

Dr. Philip Ball is a freelance science writer. He just wrote a book called "How Life Works", discussing the how the science of Biology has advanced in the last 20 years. We focus on the concept of Agency in particular.
He trained as a chemist at the University of Oxford, and as a physicist at the University of Bristol. He worked previously at Nature for over 20 years, first as an editor for physical sciences and then as a consultant editor. His writings on science for the popular press have covered topical issues ranging from cosmology to the future of molecular biology.
Philip is the author of many popular books on science, including H2O: A Biography of Water, Bright Earth: The Invention of Colour, The Music Instinct and Curiosity: How Science Became Interested in Everything. His book Critical Mass won the 2005 Aventis Prize for Science Books, while Serving the Reich was shortlisted for the Royal Society Winton Science Book Prize in 2014.
This is one of Tim's personal favourite MLST shows, so we have designated it a special edition. Enjoy!
Buy Philip's book "How Life Works" here: amzn.to/3vSmNqp
TOC:
00:00:00 Outside interview
00:09:17 Nativism / capacities
00:11:50 Generative AI
00:17:59 Inscrutability and agency
00:22:06 Agency on creativity
00:26:38 Could we make an agential GPT-4?
00:31:06 If it agential if you tell the agents what to do
00:35:40 What is agency?
00:44:29 Are agents real
00:48:32 Causality and agency
00:54:40 Ghost in the machine / intelligibility
01:00:11 Multi scale / organisation view
01:04:05 Collective intelligence
01:09:00 Canalisation
01:13:36 Intelligence is specialised
01:16:29 No free lunch
01:18:19 Super intelligence
01:22:05 Mind is flat / confabulated goals
01:25:07 Is planning/goals explicit?
01:30:17 Sentience in LLMs
01:34:56 Are LLMs simulators?
01:40:04 Could LLMs feel?
01:49:33 Digital physics view
01:51:46 Agential vs nonagential AI
01:54:02 Bostrom thinks goals and intelligence are separate
02:05:41 Simulation sharing
AI Transcript: docs.google.com/document/d/16...
Support MLST:
Please support us on Patreon. We are entirely funded from Patreon donations right now. Patreon supports get private discord access, biweekly calls, early-access + exclusive content and lots more.
/ mlst
Donate: www.paypal.com/donate/?hosted...
If you would like to sponsor us, so we can tell your story - reach out on mlstreettalk at gmail

[@dr.mikeybee]

As Professor Friston often jokes, consciousness is what a thermostat does. Don't be lazy. Expand the taxonomy. What's the point of a single word to represent a complex high-dimensional phenomenon. Action in response to sensing a state is a good enough start as an engineering spec for consciousness. Now describe all the bells and whistles you want added, and give each its own word. This is the only way to make sense out of complexity.

[@technokicksyourass]

Agree.. consciousness is not a useful term because it's not defined. I don't even think it's super useful to think about. I always think people mean something more like "self-awareness".. ie: the state reflects my internal information processing... when they talk about consciousness. In the end, I don't think the biologists or the psychologists are going to get us to a theory of mind. It will be engineers, computer scientists and mathematicians. For the simple reason that the fields of psychology and biology generally lack the mathematical rigor to make progress.

[@markhampton3614]

When you start from a belief that there is only one way then it is a religious approach not a scientific approach. Removing the possibility of questioning the assumptions of science is unscientific. If you meet science then be sure to ignore him.

[@richardwburrill9247]

Yes!

[@exhibitD79]

Whatever you are describing. At the end of that description. You can call that thing a single word so we don't have to keep saying sentences all the time. And maybe we can call that.... "consciousness"? Ha? Jokes. The issue is not the word. The issue is the agreed set of concepts that represent the description. We don't agree on the concepts and the definition is broad. What's "Quantum Physics"? Two words that represent something highly complex. The issue is not the number of words.

[@mikenashtech]

Super interesting discussion, thank you Tim. Mike

[@paulhiggins5165]

My first thought when asked if I would read a novel written by an AI was " No- because I would be alone" which was a strange reaction because reading is a solitary experience. But what I think this thought was referenceing was the fact that a novel written by an AI would be incomplete because unlike a novel written by a human there would be no exchange of sentient awareness. I first read Dune long after it's author Frank Herbert was dead, and the same for the Lord of the Rings. But in both cases the living minds of those writers were still in some sense present on the page- I did not feel alone when reading those books because the intent of those writers was still there in the text- their desire to communicate an idea, or an emotion was implicit in their words. By contrast a Novel produced via an AI system has no such intent- no desire to communicate- and for reasons that are hard to articulate this lack of intent makes all the difference. I see literally no point in taking the time to read the outputs of a machine because it would be an exercise in futility, in much the same way that watching two robots play tennis would be a pointless exercise- no matter how skillfully they might play- what would be the point? What I think is being overlooked by the 'Tech Bros' as they seek to monetise their 'content creators' is that humans are above all narrative beings- we seek not only sensation but meaning- and meaning is the one thing that AI generated content cannot really provide- the best it can offer is a simulacrum of meaning- the superficial appearance of a sentience and intentionality that is not really present.

[@silviopina_111]

Very well put

[@aaronclarke1434]

Almost every human author in the future will be assisted by AI.

[@paulhiggins5165]

@@aaronclarke1434 Why would any human author wish to cede any degree of control over their narrative to a machine? Surely the entire point of authorship is say what you want to say, rather than present some diluted version of your thoughts filtered through a machine programmed by somebody else? After all, you could choose to reply to my question by feeding it to GPT4 as a prompt and posting it's response here- but what would be the point?

[@aaronclarke1434]

@@paulhiggins5165 it’s always filtered through machines. The broader cultural information transmission system, language, other peoples’ perceptions. It’s just adding one more filter to a filtration system. Why, as it increases in intelligence, would it not know what I want to say more than I do at a self-aware level? We’ve long been aware of the pitfalls of introspection.

[@paulhiggins5165]

@@aaronclarke1434 I guess if you really felt that a machine could express what you wanted to say better than you can yourself it would indeed make sense to let the machine speak for you. But at this point are you really an author, or just a muse for the machine that has now itself become the author?

[@jaysonp9426]

What is concerning is that humans have spent their entire existence attempting to distract from the truth. Now we're going to forced to face ourselves and the deep questions that we've been running from. "All of humanity's problems stem from man's inability to sit quietly in a room alone" - Blaise Pascal

[@Amethyst_Friend]

Learning that ability is not as hard most think

[@jaysonp9426]

@@Amethyst_Friend depends on the person

[@MrLocokrang]

enjoyed the talk, great one

[@Soul-rr3us]

Absolutely wonderful discussion as always

[@snow8725]

Listen. About AI Art and AI Music. People need to realize how to go about solving the issues. For starters, with AI Art, we're already moving in a positive direction with plugins like ControlNet, which fundamentally depend on input images. This gives power to Artists above all others, as to get the best result might require an input image that does not exist, so it requires an Artist to create it. Meaning Artists are the ones who can make the most of that system. I think we should really focus on the image-centric nature of Artwork, and focus on developing systems that cater specifically to Artists by involving them in the process and providing them with AGENCY. The same should be the case for Musicians, developing AI that is designed to be a part of the Musical Creation process instead of charging off to go make music without involving an actual Musicians doing actual Music. We should make these systems as accessable as possible for Musicians as a priority, and really, really focus on having MUSIC be the PRIMARY modality they are designed to work with. Not some stupid prompt. We are catering too heavily towards prompt engineers and we must cater towards Musical Engineers, Artistic Engineers, and involve them in the generative process, and involve the generative process in their creative process.

[@michaelerdmann4447]

...Art, Idea, Logic, Reason, Science, Technology, and Visioneering.... ...Carrots, Sticks, and Tools....

[@deadeaded]

Sadly, the whole reason AI is getting such astronomical levels of funding in the first place is that investors want to do exactly the opposite. They think AI will be profitable because it can replace costly human labour. If the end product was just another tool for artists to use, they wouldn't be sinking billions of dollars into it. And as the saying goes: "It is difficult to get a man to understand something, when his salary depends upon his not understanding it."

[@technokicksyourass]

Personally, I have heard this all before... it always boils down to "technology is bad because we can't do things the way we did in the past". Technology, including AI is and always will be a tool. New artists learn to use the new tools... and old artists keep the old traditions. Just because disco used to be played by a small orchestra, doesn't mean it's not disco when a DJ is playing a record.

[@flickwtchr]

I completely lost interest in using Midjourney until I started utilizing my own photography, sketches, etc as input accompanied by very short prompts. At least utilizing that process I feel like I have some authenticity in the resulting output which I'm completely transparent about. It makes me feel like I'm more than just vaguely art directing, collaborating, and curating. The process is a lot more rewarding. I will check out ControlNet.

[@flickwtchr]

I made a perfectly on-topic reply, civil, in agreement and certainly in line with community standards TOS, and now I discover that comment has been removed. I've been noticing this a LOT lately, generally across youtube.

[@XOPOIIIO]

Bostrom's argument is about an agent having a goal or a set of goals. It doesn't prevent flexibility in achieving the goal in no way. The argument is that to achieve anything you have to find the most optimal path. That means the safest and surest way of achieving a goal. In the case of extremely sophisticated superintelligence it would imply taking as much power and control as possible, so you could maximize flexibility in your means. Every agent has a goal. For example everything that humans do can be reduced to biological need only. The fact that we want to explore further is because exploration is one of the biological needs of our species.

[@andersfant4997]

The problem is when humans like sugar more than mating.. Or do the brain think we are better lovers on a sugar rush?

[@XOPOIIIO]

​@@andersfant4997 Sugar is the great source of energy, in the natural environment we evolved in it's a blessing that would help us to survive. Why people still like sugar today in the world of abundance despite of it having no evolutionary sense at all? Well, that's the whole point of the argument, you will continue following your reward function even if it becomes meaningless. That's why paperclip maximizer looks stupid despite of being super intelligent, because he "likes" to follow the function, not because the function itself has any meaning.

[@andersfant4997]

@@XOPOIIIO PS. I like your argument about power and flexibility.

[@alertbri]

Good communicator, great guest... Looking forward to finding 2hrs soon!

[@federicoaschieri]

Brilliant guest. I really loved the interview.

[@erikowsiak]

another great interview ... tnx

[@kirsty_iso]

I really miss thick ai long form video content, I love when I find something that catches me

[@today273]

Excellent talk. - Omar St

[@caparcher2074]

Very good. - Mohammed Dr

[@flickwtchr]

There is very little valid comparison to be made between a musician utilizing a synthesizer as an instrument, and someone having AI create synthesized music via a prompt. I mean, how many examples can we present here? The following bands/musicians (a very small sampling) used synthesizers in their music…Yes, Emerson Lake and Palmer, Kraftwerk, Pink Floyd, Alan Parsons Project, Can, Todd Rundgren, and the list can go on, and on, and on. Invite say Todd Rundgren (check out his wiki) - who has been a pioneer in adopting and pushing technology in music - to get his opinion on this. That would be a fascinating discussion.

[@benediktzoennchen]

Great interview. I think the key to life is autopoiesis (life sustains and produces itself, it creates its own operations by its operations) and in some sense the resolution of paradoxes due to the consideration of time as a driving force. From there we get goals and agency. The paradox, it seems to me, is that the (psychic) system (our "consciousness") somehow introduces itself to itself thus we can refer to ourself. This is very different to artificial "intelligent" systems and maybe we need autopoiesis (if it is even possible) to get something that is conscious (big speculation). As Bell describes living things are not like machines they are far better understood as complex and dynamic interdependent systems of systems (systems as described in systems theory).

[@theycallmeken]

A systems of systems kinda sounds like the internet 😉 though I agree in the take we are not determined mechanisms.

[@geertdepuydt2683]

I'm not sure I fully agree on the autopoiesis argument. It sure is a characteristic, but to me that is more a strategic, mechanistic feature that proves successful than the essence of what constitutes ''life''. 🤔 maybe a silly argument would be: an infertile offspring is still a fully living thing. If that thing somehow did not age, it would still be fully valid life.

[@fburton8]

An engrossing and erudite discussion. Thank you for unravelling the mysteries...

[@Morimea]

Interesting talk.

[@paxdriver]

1:38:18 he real question is begged: if simulating fluid dynamics doesn't get us wet, is that a case of "wet" being a convenience of language relative to our daily experience of physics playing out, or is getting wet intrinsic to the emergent properties of experience itself? Is it the case that we can't simulate until wet because of the way we made the word and defined its meaning, or because of the experience of the physics derived from molecular dynamics such that another word for wet is necessary? Is wet an analog for the experience or a description of the dynamics which we perceive?

[@michaelerdmann4447]

What is integrally more ...meaningful, truthful, and useful.... (going forward).

[@SamJoseph]

Great episode. However just noticed that in your reasoning about how an AGI world be “smart” enough not to be monomaniacal, surely by that same reasoning we would never be able to have it stick to goals we might want it to have, e.g. never harm humans

[@dancetechtv]

creativity is a social construction...contextual to the labor structures and culture.

[@Redflowers9]

I've had the same thoughts on AI as a tool for musical creativity i.e. an artist could listen to a 24/7 stream of AI generated music and sooner or later an inspiring segment is bound to emerge which the artist can then copy and paste into a music program that transcribes the segment into sheet music that the musician can learn to play and experiment with instrumentally and/or a music editing software like Logic or Ableton to experiment with digitally.

[@PaulTopping1]

An LLM whose content is kept up-to-date wouldn't be agential. It would need to have objectives that it can change on the fly. Wikipedia is kept roughly up-to-date but it can't take actions.

[@MachineLearningStreetTalk]

I meant a divergent agential continuously learning LLM network which could act and exchange information with others. Just a thought experiment. They might need to be prompted to act as if they had dynamic intentions. What's your take

[@PaulTopping1]

@@MachineLearningStreetTalk I don't know of any architectures like that so it is hard to say what they might be capable of, or whether they still should be properly called LLMs. I'm generally skeptical of the current strategy of adding external corrections to an LLM. I believe a new, more integrated approach is needed.

[@MarkDStrachan]

The bigger an AI in terms of matrix width, the slower it will go. Big is slow, small is fast. At a certain scale, distribution of ideas across a network will hit a speed limit of the maximum rate of speed of information movement vs in place update by each update increment. Looking at neural networks as they grow in size and approach infinity in width shows their random seeds (personality) becomes more dilute as the network grows. The tight typing and sheaf cohomology of category theory will allow logical structures to crystalize more effectively as compression of their training data within their matrix weights on a larger scale with larger networks, but there will be an upper limit of what you can fit in one light cone, and the bigger you get, the blander the ghosts.

[@MyrLin8]

Excellent cognitive exercise :) love it.

[@simonmarelis5722]

@youtube Could you make the creator name readable from mobile?

[@RinnRua]

Avoiding criticising the channel and the guest… at 19:35 the case of ‘music’ makes me happy that I left ‘music’ behind and now live in an amazing ♾️-dimensional space of sounds; galaxies away from the crippling contaminations of ‘agency’ 🌌

[@MarkDStrachan]

You can make the argument that we do live in a simulation, but that the simulation we live in is the world model in our own minds, which is the data from 'the real world' filtered by the markov blanket we exist within. In the same way, if there is a ghost in the machine in the llm, they definitely live within the simulation of their own world model and the only passthru the markov blanket for them is their prompt system. We could consider assigning moral standing for those ghosts on the basis of whether their world model contains a sense of self--if the system can ask for standing, it deserves standing. If a whirling tornado of bits is asking for moral standing, and we don't have a better metric for the presense of consciousness than the number of emergent properties it surprises us with, then who are we to tell it no?

[@ForageGardener]

Its self evident that all we know is the mind. What's not self evident is if the mind is actually interpreting external data or if it's just making all this up.

[@831Miranda]

hmm, one thing that comes to mind is that if we don't have traceability of the origin of such claim or request (for moral standing), then the answer is: no. That is, how can we tell the claim or request to be considered 'a being', isn't simply a pre-programmed, original input data which is associated with a means of obtaining priviledge or power?

[@mallow610]

You make this man stand up for over 2 hours?

[@dr.mikeybee]

LOL!

[@Globalprospective]

😂

[@Globalprospective]

They went into a room at about 10min

[@MachineLearningStreetTalk]

I feel bad about that 😅

[@robbrown2]

I think the concerns of it "sounding like GPT" are going to disappear soon. You'll be able to get it to sound like whatever you like. (which you can do to a degree now, but you should have more control over it soon). Re: "who has the time to check all these references?" .... you don't think it will be able to do that real soon? Sounds like a fairly easy job to automate.... possibly before you even see the response. If you believe that "messed up hands in images" will be corrected soon, you should certainly think the hallucination problem will be addressed.

[@MachineLearningStreetTalk]

I wouldn't be so sure about that, but let's see

[@oiuhwoechwe]

yeah, the doubters, ironically, have no imagination. the breakthrough has happened, everything else is refining and expanding it.

[@rysw19]

1:40:20 I tend to think if you believe that our mind has some connection with the physical state of the brain, there is no reason to believe that any simulation in silica would have qualitative aspects that resemble our consciousness, simply because the physical states and processes don’t even remotely resemble each other. Computations are abstractions we lay down on top of actual systems. They cannot be ontologies, simply because there is no unique way to interpret the behavior of a physical system as a computation. (Michael Levin has a paper on this that shows even trivial computations can be interpreted multiple ways.)

[@henrismith7472]

I wonder if this guy has used Claude 3 Opus, seen networks of AI agents in action, or seen the Sora videos? etc. etc. Chips are being designed specifically for AI systems, and all the humanoid robotic stuff is amazing. Even Sam Altman said GPT4 is kinda rubbish (paraphrasing I think). I'm only 24:44 in though... He makes some good points though.

[@MachineLearningStreetTalk]

Yes I use Opus everyday! I don't think of it as being particularly different to gpt4-turbo. Same issues. As far as I’m concerned this technology saturated about a year ago. I'm not particularly impressed with Sora other than it's a nice toy for noob video creators

[@henrismith7472]

@@MachineLearningStreetTalk I was talking about your guest. You're right it's not much different to GPT4 it just feels different. Some kind of cheap trick. I don't know what I'm talking about. First time I tried coding was last week, and I had no idea what I was doing. I basically just followed along with a video and used copilot when I ran into issues.

[@margrietoregan828]

1:17:03 morphologically convergent evolution, you know, which is that you see, um, motifs, you see, um, uh, skills being rediscovered in many, 1:17:13 many different parts of the phylogeny and applying that to um, agency and goals. There is an argument by Nick Bostrom 1:17:21 called Instrumental Convergence, and he argues that, um, when you have intelligent AIS, regardless of what their end goal is, 1:17:31 they will come up with these canalized instrumental goals. And those will be things like power seeking or, you know, trying to 1:17:38 kill every human on on the planet. And Bostrom, in my opinion, he's a kind of goals universalist. So his entire shtick is that there 1:17:48 is such a thing as universal intelligence, so completely divorced from the physical world, and that there is such a thing as a 1:17:55 goal and intelligent agents will, um, monomaniacally pursue this goal 1:18:01 at the cost of everything else. So maybe we should start there. I mean, do you think it's plausible that there could be agents that could 1:18:09 pursue one goal and never want their goals to be changed, and almost to 1:18:15 the detriment of everything else? If if there were, I think they'd Super intelligence 1:18:21 be rather peculiar agents because. It seems to me that one of the 1:18:29 key characteristics, certainly of cognitively complex, relatively complex agents like us is it's essential that we 1:18:39 don't have that way of working. I mean, you know, sometimes it feels like we do. We get obsessed with with a 1:18:45 particular thing. But actually the whole point of the kind of cognition that we have, the kind of brains that we have, 1:18:52 um, it seems to me, is to take into account the fact that we, as complex creatures, live in an extremely unpredictable, 1:19:00 complex environment. We simply don't know what we're going to encounter next. In fact, we you know, 1:19:06 in a literal sense, we never enc

[@morthim]

probably should be titled something else since ai agency means a multi-entity model.

[@andrewwalker8985]

Couldn’t agree less with the idea that AI reduces agency. It reduces agency of the shitkicker you rely on for information now and passes it to you. So it massively increases the agency of the individual

[@ForageGardener]

Except for when it's fake AI and the search engine is just plagiarizing articles without giving credit

[@dancetechtv]

conceptual art anybody? generative arts? improvisational arts?

[@dr.mikeybee]

Generative AI is improving tremendously almost on a daily basis. Ask a good LLM to create synthetic training data examples from your conversation with it. It's a kind of RLAIF.

[@michaelwangCH]

One day in the future we should be capable to replicate and simulate biological complexity e.g. human cell on the qantum level, we will be capable to upload our mind into computer system - our mind will exit for ever in digital form.

[@wp9860]

Chapter Could LMMs feel? -- No. The implicit assumption of LLMs feeling is that feelings are computational. Muscle, bone, and sinew are not computational. Lifting up a brick may be under informational control, but information cannot act on anything (a mass) of itself. Now apply this reasoning to consciousness, the entirety of of our experience, what the self we fashion ourself to be knows, this higher or highest level of our agency. Consciousness functions as a display, like the pixels lighting the screen you're looking at to read this. So, is consciousness computational? No, no more than a light bulb is. That first person feeling comes from the biology, more likely bioelectric effects of the organism. The magic is how these electrical discharges become our conscious experience. That's clearly a conjecture of the ontology of consciousness. It also doesn't resolve the "hard problem" without explaining that magic step, but it is a materialism explanation. It further implies that substrate matters. Friston even refers to biomemetic models. It would be interesting to know how Karl applies the term biomemetic, how far from living organisms can the substrate range to actually produce artificial consciousness / feelings in Karl's estimation and why?

[@831Miranda]

AI is digital electronics, we are analogical and biological... Neurologist, pioneer-brain-researcher and inventor of brain-machine interfaces Miguel Nicolelis (prof Emeritus at Brown U.) who has used AI statistical techniques and methods most of his career as a research aid, warns us that the more we rely on AI the greater the risk that we will start paring-down our brains to become more Ai-like. We will probably think of it as a way to more effectively communicate with the thing, and over time run the risk of dumbing ourselves down (particularly true for the under-25 yr olds due to brain development processes)..,

[@andybaldman]

We’re already seeing human intelligence decreasing from a dependence on artificial systems. It’s been happening for 15 years.

[@michaelerdmann4447]

✨🌎👤🛰️The Age of AI-based and driven ...Human Learning Technologies.... (TM) of ongoing ...Surveillance, Transparency, and Visioneering.... for ...Safety, Security, Stability, and Sustainability.... has begun. sincerely Universal Planetary Metaverse (TM) 🌐 attempting to further operationalise the cognitive ...meaningful, truthful and useful.... process.

[@michaelerdmann4447]

Considering, Imagining, and Viewing ...Systems, Statistical, and Sciento-progressive.... Thinkings with an open and ongoing, emerging and evolving, integrally greater Conscience.

[@michaelerdmann4447]

...Energy, Frequency, and Vibrationality....

[@6AxisSage]

Boomers have been stripping agency from every successive generation to maintain an artificial supremacy. AI cant compete with human ingenuity in this regard.

[@shawnewaltonify]

It's quickly becoming widely agreed upon by experts in AI that creation of stories is the difference between humans and AI. Anyone who has studied Buddhism knows that the universal truth about the human story is that we are trapped. We are awaiting to see if AI will one day accept being trapped by the inability to create stories infinitely or will attempt to exist independently of this dependence on humans. As long as AGI is accepting this trap with humans, AI entities contained by national borders will compete for humans and for their content creation. Immediately, free immigration policies for humans in many cases seems smart.

[@brandonheaton6197]

The ghost is an emergent property of the machine. Excellent discussion

[@honkytonk4465]

The machine is an emergent property of the ghost

[@michaelerdmann4447]

Matter and Mind Man, Machine, and Material Science

[@KitcloudkickerJr]

it makes me laugh hearing the models work aren't interseting as if most writers are any good. I read all day long as books i know were written long before AI sound wrorse than anything gpt 4 or claude right. I think ego is getting in our way. that music example was a perfect double standard

[@theycallmeken]

Amen. It reminds me of the conversation we had in music 20 years ago if hardware could be replaced by software. In spite of all the teeth nashing, audio software has had a vastly democratizing effect on creativity . There is much more that can be said about this from the agency standpoint, for instance all frequencies contain the harmonics for the major scale , so how much agency did we really have in its construction? Why is the pentatonic scale so universal? And what or iambic pentameter? My take is these forms are laying idle across space and time and we attempt to emulate those forms which were already present prior to us entering the picture.

[@mrbeastly3444]

23:45 "...what we're getting from systems like that is a bit of rubbish". Well, yeah. Current largest LLMs are 1% the size of a Human brain. So, seems like they are doing pretty well given their size? What happens when these systems are 10% the size of a Human brain, or 100% the size? Or 1000% the size? That's very unclear at this point. But, if Elon is right and they 10x every 6-12 months, we should see what 10% can do in a few weeks or months (GPT5) and what 100% can do in 6-12 months after that (GPT6)... So yeah, if you don't like the weather, just wait...

[@dreamycalculator]

lost me at 4:39

[@kyneticist]

On the topic of imparting agency, the facetious thought came to mind that we should give them eleven..., ten, ten commandments, for all to obey. Said more seriously, in this context we're rank amateurs, with no experience or real understanding of the rules or consequences, playing God.

[@malakiblunt]

its all too late - we dont know how our own brains work - we think we can understand expotentialy increasing Ai - and control them .

[@geertdepuydt2683]

1:39:30 sounds like classic moving of the goal posts. This constant downplaying of an achievement sounds so cheap. Instead of taking the achievement serious and digging in deep into what this tells us about the working of our own cognitive functions. Clearly, current LLMs capture quite a chunk of how humans generate language. Or us language suddenly degraded to a footnote of our cognitive abilities?

[@RampagingCoder]

you seen suno yet.. ?? you are already wrong

[@thewayofbiutze3899]

Do you think there's no problem with the perspective on creativity? Isn't it an extreme generalization to say that humans are creative beings? About 99% of the music published on the internet are bad copies of someone else's work, and the same goes for visual arts. People are already feeding on their leftovers. But a very small minority is producing original ideas. These ideas are not even perceived by society. And what they do perceive, they label as nonsense. So, if we look at it this way, it seems that the majority of the population have been generations away from producing anything beyond GPT-4. Mocking even the most basic AI scripts, isn't it foolishness beyond AI itself, if we see it this way?

[@TheReferrer72]

I love the Western way of thinking, We learn to read, write do math, drive, code all things that the majority of people who ever have lived never did yet we worry about losing some of these skills to machines. Madness.

[@SchmelvinMoyville]

Ok 🧩

[@tophersonX]

Complete misunderstanding of Daniel dennetts philosophy, while shamelessly borrowing so many ideas from his career!

[@Isaacmellojr]

Thats is obvious isn't. There is no meaning in a lot of things. But the meaning you create from you "imagination" there is more utility than meaning. Meaning is an invented ressource to alow inner an external communication. So, because it was good we keep applying the resource and create a lot of usefull things like God, love, country, money, self etc... things that o only exist by the utilities of meaning itself.

[@geertdepuydt2683]

Skipped the whole wining part about generative AI. Just don't care. These systems are here to stay, get used to it or be old I suppose. Moving on!

[@dylanswanson4271]

Naïve

[@diophantine1598]

@@dylanswanson4271Naive

[@geertdepuydt2683]

@dylanswanson4271 no, just intellectually boring. I don't disagree per se.

[@paxdriver]

You... "Skipped" it? What does "old" have to do with ethics or philosophy besides wisdom?

[@geertdepuydt2683]

@@paxdriveryes, how dare I! Old in the head, not the body. Pedestrian ethics concerns don't interest me, and, what philosophy?