matthewputman

scientist, musician, author, producer, poet, father
CEO & co-founder, @nanotronics

[matthewputman @matthewputman]

Everything mysterious and beautiful about being human. Ai is not better at writing and drawing, it is different. Our voices, and creations are the things transport us. If there are parts of your life, even if it is a job, that AI does better that you do not value as being human, even if it has been necessary, then let’s welcome it while focusing our how to change economy incentives and reframe purpose.

[matthewputman @matthewputman]

For sure true about the talent and probably will be, if it isn’t already, a loss of technical superiority in AI. But why is that something that should concern everyone? If talented people find platforms that build something that is better that is a win for all of us that use AI. We are not stuck with one company. We are either stuck, or privileged (deepening on your perspective) on these technical advances not on open ai dominance. Maybe more important is to avoid the race for better LLMs and scaling alone as the only important AI strategies. I am happy to have these people finding homes that they feel empowered to do that. The point seems to be that open ai has such a huge user base that I dare ai is all they will know is temporary I think.

[matthewputman @matthewputman]

I dont mind citing AI, as it is a knowledgeable agent that provides different perspectives through a different mechanism. That makes life interesting. But the bigger point is that whether it is ai, or access to any information that you can get during discussions with a phone being asked for anything derails a pleasant and useful part of what makes anything that requires discourse function well. The journey through ideas and personal experiences produces uniquely human discoveries and bonds. We used to do this, and now if we force ourselves to not find answers, or ai opinions before exploring our ideas that are shaped through reading, arts, sciencific exploration and maybe most importantly meaning conversation , we are losing something that is so exciting to be a part of as a human. I love discussions with AI, and often we are now doing it better than discussions with other humans. We lose something valuable by not doing both.

[matthewputman @matthewputman]

Yes, this actually makes sense to me. Things as complex as Protein folding and (in the work I do) factory control are governed by underlying physical constraints with measurable objectives. AI can optimize those spaces because the reward functions are much clearer, even with systems that are more complex.

[matthewputman @matthewputman]

@adelayida210519 @GaryMarcus Funny. Most smart people know they do not understand very much , but of course I recognize your sarcasm. In this case @GaryMarcus has spent a good part of his life understanding AI, and I know a bit from my work, but what a drag if either of us understand everything.

[matthewputman @matthewputman]

@elonmusk I agree for the moment, but what is the inflection point where prices, or currency exchange in general are no longer relevant? I was encouraged by a shared view of an abundant world that exceeded UBI. Are there contemporary economists you follow when tracking this new future?

[matthewputman @matthewputman]

@GaryMarcus Yes. I said that wrong. I have a situation like this, and I should not have projected my survival methods on you. You certainly deserve attribution for your work. I apologize for suggesting otherwise. You are right to point it out.

[matthewputman @matthewputman]

@GaryMarcus Yes. Demis seeks truth and progress in ways that matter, but so does @DarioAmodei . No one has gotten it fully right, but they are honesy doing it for the right reasons and it will prove more important than those of us rhat philophoze about it.

[matthewputman @matthewputman]

You were the guest, but i think that LLms have exceeded your expectiations , but your basic thesis is still right. I think that what you said about gpt-6 not being different than gpt 3 was right was right as neuro symbolic , or adversarial RL , is still true. Do you stand by your 2020 thoughts on this, or are you surprised by the progress that has been made @GaryMarcus ? You and I worked together on some exciting applied deep learning but what would chompsy or Minsky think now?

[matthewputman @matthewputman]

I had a podcast where I was fortunate to have amazing guests. This was just as GPT 3 came out. Despite LLMs exceeding so many expectations, I think it worth listening to @GaryMarcus and keeping in mind that the need for new approaches https://t.co/e5HqTqYED1

[matthewputman @matthewputman]

I was lucky enough to see some of this come together. Check it out. Fun and original work by my daughter Juliette https://t.co/uBw85NmxGB

[matthewputman @matthewputman]

@HiFromMichaelV Thanks for sharing that blog @HiFromMichaelV . @jessi_cata wrote a great piece. I wonder if people know how many people you have influenced who now do the most important work in AI and so many other things. Cults insult people from the world while you inspire them to build it.

[matthewputman @matthewputman]

Yeah. I am not arguing with you at all. In fact I am not arguing with anyone, I just want to know where the ideas start, the politics interfere, and where the speculation is. I don’t know genomics well enough to have an informed opinion, and in the science I am not giving an opinion. Generally you start with how nature works, and figure out what to do with it as a society after. Premature speculation is my only concern, but I have so much to learn before I can decide either way on matters that are either true or not. Otherwise it is a political argument disguised as science.

[matthewputman @matthewputman]

I guees that ignoring science for consensus is real, and hinders progress, but there is something to prioritization, and judgement about what we should do on some things matter more to me more . That said, that is why different disciplines exist I guess, as some should focus more on what they are trained on without bias. I just see a time where it becomes a question of what matters in genetics as much as what is. In physics it just matters, if you are a physicist

[matthewputman @matthewputman]

@AmandaAskell @DanielleMorrill I hate seperating this too, though you could rightly separate those with a scientificaly curious world view, and those without. It is intentionality more than ability.

[matthewputman @matthewputman]

This video is in no way the “greatest clip ever.” Studying genetic variation matters for biology, but talking about intelligence is scientifically premature and socially reckless. At present, genetics explains population-level variation far better than individual cognitive traits, and no genes for intelligence differences between racial groups have been identified. I could just as easily argue that many of the Black artists I work with show greater forms of intelligence than the white technologists I know, but that would be an equally useless scientific claim. People are speaking from bias long before cold evidence exists, let alone a shared agreement on what we even mean by intelligence. I am more interested in questions like whether traits such as perfect pitch, so central to jazz, America’s great original art form, have genetic components. Until then, shouting about unfinished science only invites dangerous racial conclusions. But your comment is certainly true too. Lack of generic variability, will not lead to understanding.

[matthewputman @matthewputman]

@GaryMarcus @kimmonismus Of course it’s remotely possible. That’s not a very high bar. But cancer is already multivariant enough that if AI can make real progress there, it becomes a meaningful proof of concept, not just a remote possibility.

[matthewputman @matthewputman]

Jesus, Gary. You know I love you, but when someone like Dario aims high, I wish you’d lean in rather than mock it. You’re one of the most insightful thinkers I know, and it’s a waste to spend that brilliance telling us what can’t be done instead of helping push the frontier. And if you see a better path than LLMs, lay out the timeline . I’ll take it seriously.