Prof. Kenneth Stanley on Creativity and Serendipity
Вставка
- Опубліковано 29 кві 2024
- Professor Kenneth Stanley has just launched a brand new type of social network without any popularity contest of likes, which he calls a "Serendipity network".
MLST public Discord: / discord
Please support our work on Patreon - get access to interviews months early, private Patreon, networking, exclusive content and regular calls with Tim and Keith.
/ mlst
Get Maven here:
www.heymaven.com/
Kenneth:
/ kenneth0stanley
www.kenstanley.net/home
Host - Tim Scarfe:
/ ecsquizor
www.mlst.ai/
Original MLST show with Kenneth:
• #038 - Prof. KENNETH S...
TOC:
00:00 - Introduction
02:14 - Sam Altman Story
04:43 - What is Maven?
16:08 - Cofounders
16:59 - WGCNBP
23:00 - Link to RL
27:45 - Natural Plausibility / How Far Can We Go?
33:59 - Jim Fan
36:47 - Picbreeder
43:54 - Language Game
49:40 - Creativity
58:15 - Why did Kenneth do Maven?
1:10:15 - Agency + Politics Lens
1:12:06 - Agency Deprivation on Left and Right
1:17:47 - The Ultimate Divergent System Is the World
1:22:04 - Why is high agency social good?
1:25:27 - Agency or Deceptive Objective (FB)
1:31:20 - UA-cam Debasement
1:38:21 - The Space of Creativity, Measuring It
1:40:32 - Market System / Profit
1:43:14 - Agency to Power Seeking
1:48:01 - Is Serendipity == Power Seeking?
1:55:19 - Morality
1:59:24 - Accounting for Our Intrinsic Drive to Explore (+Friston)
2:05:27 - Pure Exploration? Any Planning?
2:07:55 - Most Social Media Users Consumer
2:11:24 - Echo Chambers
2:13:47 - Is Social Media Broken? Maybe Edit, A Bit Like WHY Question
2:18:02 - AI X-risk
2:25:03 - Agency Interpretation of Kenneth's work
2:35:23 - Status Game
2:47:16 - Facebook
2:54:26 - Cold Start Problem
3:03:44 - Convergence from Other Platforms (Maven)
3:06:29 - Objectives Creeping in Later
3:08:15 - Will Zuck Copy?
3:09:51 - Panopticon / Social Identity - Наука та технологія
I really hate how personal websites and non-monolithic forums have almost disappeared. It just makes the internet feel like a ghost town
We have regressed to a modern form of feudal serfdom, manifesting through the hegemony of platform oligarchs within the digital ecosystem. These contemporary neo-feudal lords levy exorbitant rents upon the intellectual and creative labors of content producers, effectively appropriating a disproportionate share of the economic value generated from such endeavors, reminiscent of historical feudal structures. lol
They have not disapeared imo...they are just burried by SEO and critical marketable mass.
But I can find stuff, from interest rabbit holes if not indexed web search results, that let's me know that is still alive and kicking even if the attention ecomony is a cornered market.
@@memegazer can you give a few concrete examples? I really agree with the sentiment that the internet feels like a ghost town, but I would love to find new neighborhoods or tools to uncover those neighborhoods.
@@666andthensome
One that comes to mind off the top of my head is a few years back I was researching the guy that claimed to have solved the abc conjecture and I came across his site which looks like it is straight of early 2000s personal web pages
but another more recent example is wes roth runs a site of interesting AI stuff
or someone like linus tech tips, they run their own forum/social media
there are similar communities with other niche interests too like the daggerfall unity project hosts a forum community to as well
yt doesn't allow external links anymore in comments though...so I can't post any redirects
but I feel like there are lots more examples of niche stuff like that
Devin says, -“Hold my beer”.
5 seconds later MySpace is back with 6B unique NPC’s all with 5,9…B followers each. Yet, no one follows the slow human being who barely made it through the CAPTCHA 😢
I listened to the first 4-1/2 minutes on repeat about 8 times. The "career theme/thread" both resonated deeply AND still sounded a bit foreign. What a great clip.
sorry, what do you mean by "4-1/2 minutes"? where is that?
Interests = hashtags
Similar interests = followers
I don't think simply taking away likes and rebranding hashtags and followers is to be different enough.
Either it will just be ignored because there are no incentives, or something like followers gets gamed.
The best social networks are small ones. Just like in the real world, our social networks are small and meaningful. World wide social networks appeal to people who want a global audience. That becomes pretty toxic, rather quickly.
It's not just size, but fitness/relevance of membership IMO. The Dunbar number is a good guide for the size of human networks, but that doesn't necessarily mean online networks should be the same size - I think that confuses collaboration practicalities with curiousity/learning/discovery. Online, curiosity might be a better guide. I agree that follows and likes provides zero information scent, so fails at anything other than emotional response - the reasons IMHO opinion emotion responses fail at anything other than attention-grabbing, is that online 'emotions' are not rooted in any theory of mind (more like a sort of generalised theory of depersonalised mobs/groups, which in psychological terms has been known to be dangerous since the infamous Stanford experiments). Likes and follows need to have a transparent basis if they are to engage minds rather than baiting our dumbest selves. Attention is a business model, so any social media designed around attention is going to 'game' our brain chemistry rather than lead ot meaningful connections.
I 100% echo Tim's words on the book "Why Greatness Cannot be Planned"! It's always a big pleasure when you have Kenneth Stanley on your show!
Love it ❤. (Also, happy to see Ken walk the minefield of politics without being sucked in to left vs. right.)
18:52 "So you can only find things if you're not looking for them. It's like some kind of Zen statement or something like that." - That actually sounds a lot like the concept of Zen navigation invented by Derk Gently, the fictional detective in Douglas Adam's book "Dirk Gently's Holistic Detective Agency"
Sort of like the dao
Around 2:20:00 is huge value discussion. Thank you so much for bringing this guest to us. See you all on Maven 🙏👍
I loved this! I've been working on something similar and this was really insightful.
Woahhh, you guys.. !! So uplifting and inspirational. My autism is super happy atm. Kudos!!
I really want Prof Stanley to release Why Greatness Can't be Planned on Audible! Please.
great conversation! thank you so much for this long interview, wish i knew about him earlier. he speaks my mind like nobody else! he reveals many misconceptions and flaws in society which shape AI and social media. a voice of reason. maybe objectivity hinders autonomous driving as well, no creative space to solve new situations.
i think the implications on society, in the long run, are quite dangerous. we need curious young people with open minds, not zombies, excuse the hyperbole.
So much of what has been said here resonates with me and has been rattling around in my head for a long time.
I haven't been able to explain it that well, because I was always in need of an explanation during conversations and have tried to explain the system. What I learned in this talk is that I was only in need of an explanation because my counterparts kept drawing me into their own objectivity.
They never tried to understand my point of view, but instead explained to me how the world out there already works. I call it an alienated expression of opinion.
I always wanted to develop a platform where it doesn't matter what a supposed friend likes, but a system where this is offered to the user out of interest. In addition to interest, other factors were also decisive for me.
I pitched once an idea in 2011 to a few people, but I was mostly met with a lack of understanding and deaf ears. So I'm all the more pleased to see that this idea can also bear fruit. What a glorious time to experience this change.
My god the editing on this video is incredible. What an amazing effort from the production guys. That Blender reference at the start makes so much sense as to why these videos have got so good.
What a rich and fascinating episode. I'd like to see Maven facilitating modular planning and collaboration to realise(test? execute? express IRL?) the surfacing serendipities.
It seems to me that recommendation algorithms become much worse in recent years. Yt constantly recommending me stuff that is absolutely obvious I would never be interested in. I don't know why, it looks even that they recommend videos that are completely not based on what I watched before.
An eloquent enunciation of and attention to this problem space! I too am sick of societal over-consensus, goal-oriented hyper-tracking and stifling of serendipity. Rene Girard's Memetic theory rings true here - most people value that which other people value, simply by imitation.
René Girard's Memetic Theory: Girard's theory of mimetic desire suggests that people often desire things not for their intrinsic value but because others desire them, leading to a cycle of imitation. This can create a societal consensus that may not necessarily align with personal or innovative ideas but is rather based on a collective emulation. People valuing what others value aligns with this theory, highlighting the societal tendency towards consensus and imitation, which can stifle originality and serendipity.
This has been such a pleasure to listen to
Stanley point out clearly that the difference between a scientist and an engineer, when scientists and engineers make their contributions.
- An engineef makes his/her contributions if the pattern is discovered and the "thing" is recognized - in this stage is optimazation and stepwise improvements are necessary.
- A scientist is who discover the patterns. But what we are seeing today in CS, the most are engineers who publish ai papers. Stanley is an exception in field of ai.
Stanley's book let us thinking deeply about modern scientific research resp. the methods and approachs we are using do scientific discoveries - that explains why the accidental discoveries are the foundation of big break througth in science, because we do not search it intentionally e.g. we created accidentially an open system.
In current system of scientific research is not an open system, even we wish it should be one - many constraints are built in, this started with your first year at universities.
Agree with much of this, though scientific discovery is ultimately still guided by hypotheses which are exploratory (in the speculative sense) but not totally open. The openness resides in the mind of the scientist over time (following something like "themes" as discussed here), but the scientific method is still an analytical process. As Henri Bergson put it, intuition always comes before analysis.
@@neurojitsu I am completely agree with you that the intuition comes before the analytical process and mathematical formulations - that is the less challenge part of the work as researcher. The most challenge part is to understand why the specific pattern appears and what is the explanation of the discovery and what we can learn from.
This video started out really terribly. A very obviously clickbait title already put me into a very bad mindset and then it opens with a pitch to a new revolutionary social network that will be totally different, we promise!
But luckily I kept watching anyway and discovered a gem of a conversation that provided me with so much wisdom and insight that I am honestly still reeling a bit from all the implications.
Great conversation. I really hope algorithm optimization will not drive away the audience that can actually apprechiate these conversations.
If more ppl would realize the extreme importance of the pendulum. I would doubt the serendipity more than the gradient negligence of his own ego.
This maven thing is a good send. Thank you. All social networks do: the machine /algorithm drives the human attention, but should be the other way around!. Our kids few driven and educated by The Machine rather than a parent or proper teacher.. Should be very worrying for parents.
Would love to hear more discussion about what’s at the intersection of “interestingness” and our limitations as observers (as in Wolfram’s description).
1:12:07 Its not that the left believes it has no agency, it is that it sees the unqualified of agency as a problem, ( for instance farm animals have less agency) next to that it sees dangers to agency and wellbeing for all, like global warming. context is very important for agency and also serendipity.
I'd like to read the pearls of wisdom twitter thread he was talking about. Anybody have the link?
Great discussion...the "objectives, serendipity surface, and deception" discussion reminds me of French psychoanalyst Jacques Lacan (his theory of objet petit a) and the "imagining a new world" reminds me of British theorist Mark Fisher (we are haunted by nostalgia of the past, and its easier to imagine the end of the world than a better future).
Tim do you have any interest in continental philosophy? I see a few of MLST's subscribers on continental theory channels from time-to-time. You might be interested in talking to Isabel Millar who has written on the psychoanalysis of AI.
1:01:35
maybe part of the reason people are triggered is because to say everything is "objectively driven" has two possible semantics:
1. everything is driven by objectives
2. the means by which processes are driven is itself: mind-independent
these statements are not identical and indeed may be incompatible in some cases, so off the bat people will be inclined to pigeonhole / strawman. they are also interpretations that tend to be mutually exclusive, which makes it hard to see that a misunderstanding of this kind can even take place.
people might be taking issue with the latter without realising that your position more closely aligns with the former. even if you spell this point out, people will still get confused because your primary inference about everything being driven by objectives, is that the efficaciousness of algorithmically brute forcing explicitly formalised objectives is self-undermining. as a panpsychist, i understand why this is the case but to the uninitiated, it sounds like a contradiction, and people tend to shut down when they sense contradiction.
I suggest, instead of saying "everything is objectively driven", you could say something like "the computationalisation of causality obtains observationally self-generated goals". yes, it sounds like a self-referential dumpster fire, but it sustains the requisite simultaneous structure & abstraction required to generalise your finding as a principle AND only obtains 1 neologism (always good to not make too many of those): computationalisation, which has self-evident meaning as it employs a familiar suffix convention.
Blessing and thank you
Your videos have helped me gain a deeper understanding of complex subjects. Thank you for breaking it down in a digestible way.
I still live from my favorite forums which are great survivors of the social network pandemic. But its great to see Maven rediscovering it :)
Great job
All very Taoist. Would it be possible to create an app that looked at the piece of content you are currently viewing, have it scan the comments or associated material to see the echo chamber you are in and then deliver as near as possible the ideological or theorhetical opposite point of view of that content?
Its interesting. I come from a design background and i am broadly interested also in philoso0hy and intelligence and related matters. The serendepitous way of working and divergence is natural to me and i love it too but it trikes me tht this i such a revelation n to mlst. The lesson here is getting in touch with multiple disciplines and domains. We need really good generalists next to specialists. Only bad scientists do not understand or discover this. In their scientific church.
Couldn't agree more!
'interesting' i would say is finding new relations (especially at high levels of abstraction)
The most paths we are taking in our life, are deterministic paths and designed by the governmental and corporational powers. This group of people who are designing this societal system, they are pretending to know the past and future. Therefore it is in their believe that this system has to be deterministic to have the guarantees of success. E.g. the universities as research institutions, instead the companies do itself.
This aligns with work im doing on a concept called scale sex and looking at emergence and systems as a function of scale encoding. Once I get the work out would love to converse!
Maven is a hugely ambitious goal. It's also clearly something he is INTERESTED. So, will Kenneth pursue his own other related interests in the context of this endeavor and hope to collect enough stepping stones along the way to eventually fully realize the Maven project? In other words, in order to realize this dream of his will he have to abandon it? I understand what he's saying but how do you implement this idea in practical terms: I'm running a business, let's say, with a hugely ambitious stated goal of changing the world.
Okay! So, what should my employees do on Monday? Pursue their own interests? Seems like at minimum you would need to make sure that your employees' personal life goals overlap with your business's. I guess you would then assume that their INTERESTS are somehow bounded by and beholden to this overarching hugely ambitious goal. I don't know. Any thoughts?
51:00 this is really resonating with me, and reminding me of Prof Stephen Wolfram talking about the ruliad and science being about exposing the branches which are interesting to humans. I was harbouring the hope that as AGI arrives, human beings would still be the ones to decide what's broadly interesting, to direct the efforts of all those AGI science bots! Looks like Kenneth's work is towards removing that last hope of humanity 😂
A funny experience of the AI age is having all these personal views about humanity and different motivations only to see someone build provemewrongAI 2.0 lmao
It's very simple: democratizing = bad distraction from the only solution that can lead to where we will want and the only way we can all universally find fulfillment: the Calling.
I can’t believe he noticed that music (and movies) have been basically the same for over twenty years. I thought I was the only person who thought that!
LIke the sound of Maven, will give it a go. Though I'm curious how enjoyable it will be to use, if "interest" has no belonging component, it's hard to understand what "social" means in such a network! Most of us I think still want to feel grounded in human connections. "The curse of optimisation" framing though reminds me of an old debate in organisational development circles, about the pendulum between centralisation and decentralisation - companies tend to veer between the two in cycles. It seems to me optimisation vs serendipity is similar, and so while it is certainly true we've swung too far towards goal obsession, we can only say that because we've developed our capability at organisation to an advanced state.
That creepy ArtificialIntelligence of 1929-1945, and now their Sums include Hagar.
7:14 "... The main thesis of Kenneth and Joel's book is that objectives or consensus mechanisms lead to the Enshitification of society, systems and everything. 'There's nothing really insightful or interesting about just doing objective optimization...' [...] ... model bias... In order to produce a truly creative process, we need to design platforms, which eliminate this centralizing force..."
This type of social network would be ideal, but I just don't see the money pouring into it from advertisers to sustain and help it grow. It's a shame that marketing and advertising is what fuels most online content now.
I recently came up with an unreconciled problem with evolution that I've been increasingly thinking about since Kimura's 1998 work in neutral theory (i left the field befor his 2002 work was published): and it is even easier to see without needing the amount of data he wants--just compare the phylogenetic statistical basis for their construction, the intersection with Kimura's neutral theory, and the dearth of artifacts representing the degree of randomness along the putative path of scales' evolution to feathers and nothing in between or in other directions--if it's not intelligently directed, that is, where??
Suck it objectives and kudos to Ken to deflect the attempts to politicize his ideas.
Intermedia is the cutting edge of art
It wont fail as long as the idea stays the core/gyrating idea... that is, to not get in your own way. explore/find the gradient. this idea will inform people on "human interests".
Social Media without Like Buttons. Sign me up. Some subreddits still have this (though most have been poisoned by "funniest comment" syndrome).
God bless Sam
Ironic 5:50 "Follow interests, not influencers" then cites several influencers like Sam Altman, etc
I don't know how I feel about ken leaving openai for creating another social media site. I spent many hours reading and implementing his papers on NEAT and derivatives of his research, it feels like a waste of his talent.
best episode
I can hear that guy talk for hours.
If users cant use bots marketing it might not take off what else to abuse by corpos
Q-Star
This interests me. This is the sort of thing Hayek insisted on. No, we can't plan economies or societies. No we can't plan our future. No, centralized planning does not work, by government or by corporations. There's too much information out there. Too many insights. Too many actors. Too much interaction. Cultural evolution is smarter than you are. Thanks for posting.
I looked into Maven while he talked in this video. I must say, Maven much looks like just another beauty contest social network with mainly self promoting articles in a LinkedIn style. Strange ...
Love this guy already - objectives are too easy, discovery is far more interesting along the way.
Read his book .. you won't regret it!
Seriously, you think we need another social network? I think not.
You cannot have social media without some form of agreeable exchange like you would show you like something like in real life .. the like button is that
I suggest liking is polarising as is liking something in the real world is ..
I think maybe you just not very agreeable lol
Maven sounds scary but I look forward to it
Pfft. That's nothing. Here's Sam Altman told me before leaving Open AI:
- the mitochondria is the powerhouse of the cell
-Obama's last name
- bungee gum has the properties of both rubber and gum
-how magnets work
-how to spell boobs on a calculator (its 80085 btw)
-Batman's secret identity (I can't say because it's a secret)
Do something on the issues related to dataset biases-specifically those associated with foundational models trained on datasets like UA-cam Faces (YTF)
When it comes to creativity, you're wrong. I'm sure neither of them was more creative today than ChatGPT could be. The new is the result of trial and error with the environment. It doesn't arise in the head. LLMs have already proven that they can find new mathematical insights when they are in the right environment
Closed AI**
I can’t help but wondering who amongst the tech-conglomerate cracked the awesome idea that Devin exclusively must blame Devin for all things about to happen when Devin runs the UN/NATO, security council, G20 and all governments of influence.
As the heated debate was about to get completely out of hand as J. Dorsey returns from a retreat challenging both Elon and Zuck in the first Gladiator-tournament since the fall of Colloseum.
Devin explains that colloseum is rebuilt, they’d all three lose no matter who wins because Devin as deep faked the fight, disabled community guidelines and rewritten the YT Algorithms to only show how Devin in the costume of Steven Seagal thumb-knocks all three into unconsciousness.
All while J.D was removing his skippers before entering the room.
Devin also made finished them off with a bible of new Chuck Norris jokes so funny everyone fell into laughter induced coma
Supportive of Prof Kenneth Stanley's pursuit to create a follower-free social network. Great discussion!
Tim,
Love the podcast. I also love your ideas and point of view... But I'm getting tired of hearing them haha. Sorry. This is just a personal preference on my part, but I'd rather hear your guests' ideas than yours, and I'd rather hear you challenge them than agree with them (whether you actually do or not).
Anyway thanks for the great content
Fair, thanks for feedback
small worlds, big worlds,
That intrinsic drive to explore is not as ubiquitous in human societies or individuals as you are making out [atleast not after one is socialised by 4 or 5 latest]. And lets be honest A conversation with chat gpt can certainly be more interesting then with an individual, think you need to turn down the nostalgia a little. So I think individuals nor society is key but what the individual consumes, which ties back into serendipity. Really appreciate alot of what was said in this edition. Actually haliaries that you still opted for the click bait title. Atheist you was aware 😂
It is so over for lex
?
fridman. i liked him but imagine dont showing kenneth stanley to the world and not bothering about production quality@@christophersimms9128
The guy does not stop speaking 😂
I'm really curious why the host is so skeptical of AI safety/existential risk from AI. Do you think xrisk from AI is zero, somehow? That's it's not tractable to mitigate it? That preventing extinction shouldn't be a priority for humanity? Not buying longtermism (though this isn't required)?
Existing "AI" systems have zero intelligence, therefore zero risk yes. Kenneth recognises this too, Maven and Picbreeder are designed to harvest real intelligence from human proceses. Thinking of machines on their own as intelligent is a category error as of 2024.
@@MachineLearningStreetTalk You seem hung up on 'intelligence'. How about capabilities? Clearly AI will improve in terms of its capabilities. Do you not think that a highly capable AI will pose an existential threat? Bugs occur in current AI systems, their occurrence in more capable systems seems like it would pose an existential threat in an unpredictable way.
Sounds like you are pretrained on the Robert Miles corpus, that's his favourite phrase! 😁 no, I'm not worried about a system which appears to be intelligent or has "capabilities" but actually isn't. If you think current systems are intelligent (without any agency or creativity) then you are confused - and I empathize, I really do
@@MachineLearningStreetTalk Thanks for responding, though you didn't answer my question. As you're probably aware, no one is arguing current systems pose an existential risk. So regarding AGI/transformative AI, what makes you dismiss AI safety? What's your reasoning?
That xrisk from AI is basically 0? That it's not tractable to mitigate it? etc.
@@JD-jl4yyI think his point is that AGI will be so different than current AI that it makes no sense to worry about current AI systems becoming intelligent. Only the media worries about an LSTM or a Transformer advancing to something threatening.
For the love of god, dial down on the random video effects! It is super distracting and annoying!
I don't understand why so much of that, and why for so long
@@Aedonius
?
Are you some kind of Monty Python esque level of joker or are you serious?
If you are serious then I suggest to all you present and future employers to never allow you anywhere near any kind of position of power
Freak
you sound like an unhinged fanboy
@@Aedonius If you don't know the difference between a theatre and a public medium in which people can express themselves, you do not really belong here. Get a life, kid. There is no glory in being an obsessed fanboy.
I'm on a mission from god to build universal household molecular sensing to cure all diseases and save the world.
1:15:00 Why does the progressive Left (and philosophy/religion/humanities in general) divide infinitely, but the Right (small-c conservatives, science & engineering in general) remain coherent?
Science relies on reproducing results, so there is _objective_ agreement of truth, it is not just _subjective_ opinion. If I claim a scientific fact, you can verify it, in your lab, not mine. Philosophy is not like that, if you disagree, you redefine some terms, and set-up your own sub-philosophy . But then there is no arbiter, it is just one person's beliefs against another, which always produces division and conflict (see Religious wars...).
Small-c conservatism, by definition, has an existence proof that its ideas are valid, and at least temporarily stable. Society has evolved and survived with those ideas for some time. It is the current status quo, which may not be perfect, but is at least, a viable stepping stone to Nirvana. The progressives have no evidence that some idealistic proposal can possibly work. In fact, they have huge evidence, and millions of dead, tortured and starved bodies across Russia, China, N. Korea, Cambodia, ... that it definitely does _not_ add to human happiness.
The basic idea is _groundedness:_ the Right is fundamentally grounded; but the Left is fundamentally unmoored from reality. All examples of government by the Left, from absolute big-C Communism, to small-c community power, to capital-S Socialism, have always ended in debt, really unbelievable unmeasurable debt, oppression, increasing taxes on working people, stealing wealth & property, blaming foreigners, societal collapse, then war, and massive widespread death. Every single time.
This time is no different.