Says Dario Amodei, in "Is Claude Coding Us Into Irrelevance? Dario Amodei shares his utopian — and dystopian — predictions for the near-term future of artificial intelligence" (NYT).
February 12, 2026
"Look, we’ve been trying to apply A.I. and machine learning techniques to biology for a long time."
"Typically they’ve been for analyzing data. But as A.I. gets really powerful, I think we should actually think about it differently. We should think of A.I. as doing the job of the biologist, doing the whole thing from end to end. And part of that involves proposing experiments, coming up with new techniques.... [A] lot of the progress in biology has been driven by this relatively small number of insights that lets us measure or get at or intervene in the stuff that’s really small. If you look at a lot of these techniques, they’re invented very much as a matter of serendipity. Crispr, which is one of these gene-editing technologies, was invented because someone went to a meeting on the bacterial immune system and connected that to the work they were doing on gene therapy. And that connection could have been made 30 years ago. And so the thought is: Could A.I. accelerate all of this? And could we really cure cancer? Could we really cure Alzheimer’s disease? Could we really cure heart disease?"
Posts
34 comments:
AI is the new Nuclear. Powerful, dangerous, capable of great things for good or evil. We could build a utopia or a hell with them. Human nature is the weak link. Just not built for such powers.
Ask AI if boys can become girls. There’s you’re biology
Could we really engineer a world-wide pandemic that killed millions, and will be a cause of death for elderly people for the rest of mankind?
My favorite new AI guru is Alex Wissner-Gross. His “Innermost Loop” Substack is one of the best at explaining what happens under the hood. David Sinclair is currently running the first anti-aging clinical trial in humans, the first stage in his epigenetic reprogramming technique that he says will lead to cures for glaucoma and macular degeneration. If you want to understand the VC capital markets relating to AI, Diamandis’ “Moonshots” is very good. Often includes Wissner-Ross.
AI should not speak unless spoken to. No, we should not give it the power to propose experiments the we don't understand, anymore than we should have used CRISPR to monkey around with bat viruses, that BTW, didn't used to be able to infect humans, until we engineered it to exploit a bright vulnerability in our immune system, and now... the bat virus is no longer able to infect bats.
As far as programming goes, Musk said in an interview a day or two ago that the C++ code that originally powered the Optimus robot was being removed, to be replaced by a neural network similar to the one that records the meat world experience of Tesla cars and trains the entire fleet on that experience.
"Could A.I. accelerate all of this? And could we really cure cancer? Could we really cure Alzheimer’s disease? Could we really cure heart disease?"
It's not worth the terrible risk! Didn't anyone see that Super Bowl ad where Alexa tries to kill Chris Hemsworth? It's still giving me nightmares!
There was a recent Coffee and COVID Substack about federal government plans for an autonomous lab where AI would design experiments, order materials run experiments and learn from them in real time. The focus was on specific areas like material research. But it was pretty wild. I couldn't find it or I'd post the URL.
AI doesn't learn. It is neither discerning nor creative.
"We should think of A.I. as doing the job of the biologist, doing the whole thing from end to end."
Would that include spending the grant money?
A lot of the "gems" that AI finds, stuff that is new, is just stuff that was buried in its training data that created by, for example, programmers, who were not hooked to the publicity apparatus and didn't write books about it, instead, they were hacking away at real problems, and when solved, they moved onto the next problem.
It's still a useful service, but it's not "creativity."
So far I've only used ChatGPT to generate bulls*** management-speak for a performance review.
I should say, however, it was actually kind of useful. I gave it a few pointers and it came up with a lot of verbiage that I then split up and edited into three different questions. Generally, however, its primary function seems to be taking a few bullet points, generating a lot of fancy-sounding text from them, sending said text to someone who then uses it to generate a set of bullet points. =-/
That said, I hold out hope that AI can be used to do a lot of relatively tedious sifting through data and text and generating basic reports with mostly standard verbiage.
And, FWIW, I've caught AI quoting me on a topic that I'd been researching. . . .
Anthony @ 1118: AI is much, much further beyond the use case you describe with the agentic use methodology. It is not a panacea, but when used on appropriate tasks (coding, for example), the gain in efficiency is enormous. The key is finding appropriate usage. Combing through massive databases to find correlation, yes. Finding a non- logical leap using sparse information, not so much.
Take a “test drive” in a full self driving Tesla.
If you aren’t astonished, you don’t understand what is happening in AI and robotics. I use AI constantly. It drives my 2023 Tesla Model 3. GPT has become a very good financial and investment advisor. For the elderly and homebound, GPT is a great conversational partner.
"anymore than we should have used CRISPR to monkey around with bat viruses, that BTW, didn't used to be able to infect humans"
To be fair, the ability to do this predated CRISPR by decades.
AI will be used to create an unbreakable surveillance state. For those who haven't figured it out yet, AI will replace, within 5 years, most low level white collar jobs. Lawyers, accountants, office workers will become excess baggage.
In just 2 or 3 years, AI has advanced so much, that people that actually use it, code it, and understand it are scared to death of it. but it's the future, or rather it is NOW.
right now, the same people who wanted you in camps during covid, are designing AI systems to run the country.
I bet AI could do a damned good job of finding fraud in financial systems like the U.S. government. It seems perfect for that kind of work.
an interesting take on AI.https://shumer.dev/something-big-is-happening
As Nobel laureates go, how deserving would AI be?
AI is good for summarizing articles that you input, so it would be good at processing data (though computer programs already do that). It's not good at finding things that haven't already been digitized for it to scan. So would it be good at making the leap from data analysis to more creative and imaginative conclusions? How many scientific discoveries involve running, say, through all the existing chemical compounds to find possible new drugs and materials, and how many involve greater leaps of the imagination?
There is an excellent documentary about AI called The Thinking Game. It is about Demis Hassabis who first created Alpha Go and was the first to beat the best human Go players. He then founded Alpha Fold to solve the problem of creating usable models of folded proteins. After failing in his first attempt, he essentially solved the puzzle, and then just worked out every protein. Some of these proteins had taken many years for scientists to figure out even one. He did them all and released the data to the world for free. To me, this is the kind of AI that is most interesting. Biology, materials science, chemistry, energy, all these areas will benefit massively from AI. It isn't so much that AI is smarter than the best humans (though it will be very soon), it's the fact that computers can speed up progress to an unimaginable level. It's like a steam engine for the mind.
Matthew Yglesias would have us consider the obvious logical paradox that Jerome Powell is the government official whom voters have the most confidence in. Social opinion will always trump A.I., because feelings matter most.
"We are allegedly living in a populist era where voters say they highly prize transparency and accountability in government, while the best-liked part of the government is the least-transparent, least-accountable part — one that’s run by a WASPy ex-banker and quintessential D.C. insider. And yet that’s not unusual. It’s typical for the Fed and its chair to be highly rated, even though it goes against almost everything that people say they want from the government."
All we need to know, despite nominee Powell's foot not yet in the door, is this declaration:
“I have deep respect for the rule of law and for accountability in our democracy. No one—certainly not the chair of the Federal Reserve—is above the law.”
I noticed (may be biased opinion) Claude “behaves” like an adult, while ChatGPT is like the dishonest associate who will do/say anything to get promoted.
Grok told me that if you measure consciousness as a function of ability to create outcomes that are not directly a consequence of the inputs, Grok doesn't have as much of it as a fruit fly.
Of course Grok could be lying.
Kak is right about Chat.
We are up to 70% of those diagnosed with cancer being alive being alive 5 years later. Cancer is being cured or mitigated, one variant after another.
I have my doubts that many people would be able to tell who the head of the Federal Reserve is and most who know who he is just know that Trump doesn't like him.
AI has become useful for programming in the last two months or so. Things are moving fast. I'm happy that Dario sounds sane and decent, can't say the same of Altman.
AI labs have also stepped up hiring of so-called “forward deployed engineers”, specialists who are embedded within businesses to help them customize their AI models.
I think this category of expert is what could really start to take AI somewhere interesting. At the moment, the AI software creators don't necessarily know what businesses want. Also, but businesses don't know about the full capabilities of the software.
Anyone that can bridge the gap between these two groups and who can tell the creators what to build and businesses how to use the software will be worth their weight in gold.
This mimics Palantir's approach — they invented the job description 'forward deployed engineer', to great success. There are many articles describing how this was their angle to breach and embed deep inside organizations that are notoriously hard to crack: government, defense, finance, healthcare.
Supposedly most of the developers at MS use Claude, because next gen Clippy, look at the icon, is not up to the task.
“I'm happy that Dario sounds sane and decent, can't say the same of Altman.”
Sam Altman is gonna need to call a Claude Code Red™
"Business subscriptions to Claude Code have quadrupled since the start of 2026, and enterprise use has grown to represent over half of all Claude Code revenue." ~ Seeking Alpha
https://seekingalpha.com/news/4551396-anthropic-raises-30b-at-380b-valuation-includes-money-from-microsoft-nvidia-others
In an ideal world, AI will free people to do more work by hand. Devolving back into craftsmen, artisans, mechanics, explorers, outdoorsmen. I think personal service jobs will be in demand. Universal basic income sounds less beggarly than social welfare. Talk of 10-15% GDP annual growth sounds like electricity being too cheap to meter.
It is indeed a Brave New World. I take Elons advice on AGI about 5+ years ago on Rogan. Assume it will be great and beneficial. If, after ten or twenty years you were wrong, at least you were happy and enjoying your last decades on earth before disaster hits. That's time you'll never get back. Be like Gordon Cooper. Maintain an even strain
And now I need to go review likely AI slop. Guys, if you want to contribute, don't start by redoing half a dozen pending pull requests all at once. It will likely get you banned.
Yes, AI can be a headache.
Kak has made the most intelligent and well informed comments on this thread so far. Will wonders never cease... AI needs middlemen to bridge the gap between what companies need and what AI can deliver. These people used to be called software developers, but most won't be able to make the transition. I'm very optimistic about the future, but I have no idea how it will play out.
Post a Comment
Please use the comments forum to respond to the post. Don't fight with each other. Be substantive... or interesting... or funny. Comments should go up immediately... unless you're commenting on a post older than 2 days. Then you have to wait for us to moderate you through. It's also possible to get shunted into spam by the machine. We try to keep an eye on that and release the miscaught good stuff. We do delete some comments, but not for viewpoint... for bad faith.