pfisherman

Being easily verifiable does not elevate something to the level of being a grand challenge. It really drives home the point that biology is not math or coding.

I guess to me the grandiosity here feels fake and unearned - like they vibe slopped it together without much input in the way of deep thought or expertise.

For example, synthesize arbitrary dna sequences > 3000 nt in length with error rate < 0.001 at > 95% purity. Easily verifiable, beyond the frontier, and generally useful.

One mechanism by which the US encourages free market capitalism is requiring reciprocity. China is not evil - they are just self interested - and it is the responsibility of the US government to negotiate a trade agreement that benefits its citizens.

Biology already has numerous “Millennium Problems” and the rewards are much more than $1M. For example, reverse Alzheimer’s Disease. Or less ambitious develop high fidelity in vitro and animal models of human disease.

Do US companies have free access to Chinese markets?

I don’t know. Seems like a distorted sense of justice to me. I am not so sure that academic fraud is in the same league as committing felonies (like E Holmes). Also don’t know that anything beyond professional sanction is deserved for professional misconduct (that is not breaking any laws or endangering people).

Perhaps their real crimes were not having better PR, or enough money to hire a crisis management firm. Whatever it was, it really seems to have brought out the worst (in) people.

Many of the comments here in this thread are absolutely vile. Very few people are undeserving of compassion.

I think some of the people in this thread should look themselves in the mirror and ask why feel different in this case than similar cases like Dan Ariely, Amy Cuddy, or Marc Tessier Levigne.

Nice! Thank you for doing this. I am not surprised PCA holds up. It comes with some pretty strong theoretical guarantees. The problem with it has always been scalability. The O(n^3) complexity makes it infeasible to use for massive datasets.

That is where something like Matryoshka embeddings has appeal. You trade a little bit of performance for a guarantee of training + validation set coverage.

There are other reasons. One is for career development of your researchers. Another is to be good citizens, in good standing, in the community of scientific researchers. Another is to flex on people and buy class and respectability - to comport oneself as the “old money” does.

I think land value tax is misguided. Issue is what happens to retirees on a fixed income when value of their property increases? Are they forced from their home to lower priced area? Regardless of how one might feel about the morality of that it is also worth asking whether this might create perverse incentives to become landlords?

My understanding is that an ideally priced IPO should not move much from the opening price in the near term. If it pops it means they left money on the table. If it drops, then I am not sure what the implication is exactly?

Now I think SpaceX is massively overhyped, but is the share price returning to IPO opening not just a sign that the banks accurately estimated something?

You could drop the verb clause? This would make the headline accurate while keeping it punchy.

> why Americans switch to soy

And

> why developers switch to codeberg

But the cynic in me thinks that the form of the headline that drastically overstates the the phenomenon in question by implication is something that has been workshopped and is commonly used because it turns something kind of boring into a spectacle.

Same experience with Fable. Utterly useless for anything bio related.

Author of the post here wrote Salmon, which is a widely used bioinformatic tool in molecular biology. And the irony is that Anthropic has probably packaged Salmon as a tool in their Claude for Science suite with now remuneration or recognition for the original author.

There has been a lot of recent bs going on in biomed ML with companies publishing without releasing source code, restrictive licenses, etc; which have always given off a whiff of bad citizenship - Ark Institute and Deep Mind I am looking at you - but I feel like this is taking it to a new level.

Leveraging open source bioinformatics code and published methods to take in profit selling into the biotech vertical while restricting access to Fable feels downright cancerous. I think the EA crowd at Anthropic probably has good intentions, but has galaxy brained themselves into becoming bad actors that make Sam Altman and OpenAI look like a paragon of trustworthiness in comparison.

One quick piece of semantic and linguistic housekeeping for the commenters…

Heritable != Molecular / Genetic Mechanism

There is a conflation of these terms in popular discourse that does a disservice to the field of statistical genetics, imo. There are mechanisms of inheritance that operate various length / time scales other than that of biological macromolecules. For example, if you tell me what language your parents natively speak I can tell you your primary language with >90% accuracy.

So before we start getting 3 replies deep into any thead, please remember that retrospective observational data measured with unqualified instruments is notoriously confounded and that we can barely infer causal structure in controlled functional genomics experiments (much less a GWAS of phewas). So let’s all please keep an open mind and not be so certain about our beliefs.

In some ways domain experience can be a hindrance, with ingrained pathways and practices shaped by constraints that no longer apply. My personal opinion is that you probably want a mix of domain experts who are enthusiastic about AI and some kids who are free of preexisting dogma, and are willing challenge assumptions and try out things that the old heads might chafe at.

An example from software engineering is that all production code should undergo meticulous human review. Saying “no” to this sounds crazy to an experienced SWE, but might not actually be that crazy.

I don’t know about this. After some time sitting with it, I think that mid level and senior ICs - especially those slow to adapt - are going to be at risk of getting replaced by entry level “AI native” kids. Net on net it probably washes out to “normal” patterns of turnover and hiring once things settle.

Think “Smithers, we need to hire some of these kids who know computers!” Only fast forward about 30 years and str.replace(“computers”,”agents”).

The comment above is on to something. I find CarPlay to much more valuable and much more of a lock in to the iPhone than Siri. I do not think I could ever go back to using the infotainment systems that ship with cars. So makes sense why they might prioritize over Siri. And in the context of CarPlay, the simplicity of Siri is nice. I really only need it to execute a few simple commands like looking up directions, making calls, reading / sending texts, playing a podcast, etc.

No you would not expect all efforts to target AB to work. You can have the right target and the wrong molecule. Look at GLP1s as an example. Multiple pharma companies tried GLP1 receptor agonists for more than a decade and it was not until recently that Novo and Lilly got them to work.

I don’t quite follow the reasoning behind this being “unbiased”. Studies must be designed to answer specific research question. What is the question being asked by the study in question. If the question being asked is whether AB the right target (in humans), then it would need to consider other forms of human evidence, such as genetic evidence, alongside alternative explanations of failure — bad molecule, wrong isoform, bad cohort selection, etc.

Nice work! Here is an article you may find helpful if you have not already come across it.[0]. You may also want to consider benchmarking against some non ML methods.[1]

0. https://pubmed.ncbi.nlm.nih.gov/35318324/

1. https://www.nature.com/articles/s41586-023-06127-z

This is going to catch some heat, but what if the most important professional “developer skill” to learn or improve is how to effectively use coding agents?

I saw something similar in ML when neural nets came around. The whole “stack moar layerz” thing is a meme, but it was a real sentiment about newer entrants into the field not learning anything about ML theory or best practices. As it turns out, neural nets “won” and using them effectively required development and acquisition of some new domain knowledge and best practices. And the kids are ok. The people who scoffed at neural nets and never got up to speed not so much.

Edit: as an aside, I have learned plenty from reviewing coding agent generated implementations of various algorithms or methods.

I think Peter Thiel is smart, but exhibiting one of smart people’s most common modes of failure, overestimating one’s ability while not maintaining a healthy sense of skepticism about the correctness of one’s own beliefs.

Put simply, he (and many other tech bros) have galaxy brained themselves into some very stupid stuff.

Yeah, you are not wrong. The topic is a bit like troll bait for me. Probably because I have a first hand view of how the current strain of anti intellectualism and resulting policy in the US is destroying jobs and eroding competitive advantage. My observation is that this type of rhetoric tends to be produced and consumed by “elites”, and is often used to advocate for policy that limits socioeconomic mobility.

The irony is that in limiting mobility and competition from the “non elite” out-groups to preserve status, they end up shrinking the overall size of the pie.

Yes, and they are currently cleaning our clock when it comes to global competition in science and technology.

We live in a globalized economy. Rapid transport of people, goods, and information necessitates it. The high paying STEM jobs will go to wherever there is an abundance of talent, and the network effects are quite significant.

I associate this phrase with losers and people trying to sabotage the US. You know who is not wringing their hands about “elite overproduction”? China, who are pumping out tons of smart and capable STEM PhDs, and have in a relatively short time caught up to and in some cases surpassed the US in production of scientific output and technology.

This is very easy to explain. Anthropic outlines some limitations in their terms of service. Palantir accepted those terms. The DoD did not.

OpenAI claims their terms of service for DoD contain the same limitations as Anthropics proposed service agreement. Anthropic claims that this is untrue.

Now given that (a) the DoD terminated their deal with Anthropic, (b) stated that they terminated because Anthropic refused modify their terms of service, and (c) then signed a deal with openAI; I am inclined to believe that there is in fact a substantial difference between the terms of service offered by Anthropic and OpenAI.

Would this actually return memories and context? How could you know if parts or all of it were hallucinated?

I can see how being able to bring your chats with you would be appealing. But the truth is that context rot is real, context management is everything, and more often than not stating from a blank slate yields the best results.

That being said, if you have a library of images or some other collection artifacts / assets indexed on their servers that is a different story.

Why not use Claude Code from the cli and follow along in your IDE? I did not quite believe when people were telling me or understand what I was missing until I tried it, but after trying that set up I am convinced that it is superior. I don’t have any hard data to back it up, but it feels much more capable that way.

Looks like the statistical geneticists have jumped the shark with this one. This big problem here is that their endpoint (chills) is poorly defined, reported by subjects (and thus highly subjective), and not measured using any type of validated instrument. So I question whether they might be fitting a model to noise here.

In the land of drug development patient reported outcomes, even when captured with meticulously designed instruments in prospectively designed clinical trails, are notorious for being noisy and confounded by the placebo effect.