Engineer of stuff that is alive or computes. Follow us all to that blue place.
Joined January 2021
- Tweets1.4K
- Following3.9K
- Followers3K
- Likes714
Tom Knight -- e/🦀🌴 retweeted
I've been prototyping an automated peer review project. With current models you get human level peer review for about $0.10 of tokens. For $5 you can rederive every result, check every statistical inference, and audit every paper cited in both directions.
There are about 100m papers in the scientific corpus. Most are not important. Many are sloppy. For structural reasons it's essentially impossible to change published work, even to correct a typo.
For $500m today you can thoroughly audit every paper ever written. Every p-hacked result, every plagiarist, every fabrication, every citation club, everything.
I don't have $500m so instead I build a tree, focusing on the 1% of papers that carry 99% of the signal, doing a cheap first pass looking for obvious problems. I can find 90% of the nonsense for a few hundred dollars of tokens.
Any guesses where the biggest surprises will be found?
Tom Knight -- e/🦀🌴 retweeted
🤖 My latest post proposes AI safety measures that may sound less insane now than earlier this year. No one else is saying this, but many are thinking it:
Tom Knight -- e/🦀🌴 retweeted
Or just use Clojure. You don't need hot-reloading tools, the language itself allows you to incrementally build a running program, like any good Lisp. 🔥
Your periodic reminder <github.com/JetBrains/JetBrai…> exists, making the JVM likely the best combination of runtime performance & hot code-reloading available today.
Tom Knight -- e/🦀🌴 retweeted
I hope the Irkutsk event can be contained - it’s a clear example of why physical biosecurity is the biosecurity that matters. You can screen all the DNA synthesis you want but the best defense is to detect, treat, and eventually eradicate the bad stuff that’s already out there
Tom Knight -- e/🦀🌴 retweeted
Yeah, this is the whole point about fundamentally needing to 1000x change our national posture on biointelligence collection - we need technically verified data from trusted sources - world has changed
If this were a cyber event - there would be 1000 companies and govt agencies all over it -
seems so obvious
Wouldn’t be waiting in the FSB
Doubt seeds fear and bad decisions
Tom Knight -- e/🦀🌴 retweeted
We need to be monitoring every port of entry on the planet.... yesterday.
If not, we’re putting our whole biosecurity bet in situations like this on:
- the remoteness of this facility in an interconnected world
- the competency and care of the FSB
We need global collection and monitoring immediately so we can track this if (as) it starts moving out of Siberia (or every time something like this happens)
🚨Breaking. There's been a pneumonic plague outbreak at a BSL-3 lab in Siberia, Russia.🚨
"Nearly 200 People Under Observation After Irkutsk Lab Worker Dies From Plague" - Moscow Times
This is fact, and also the inciting incident in my book BIOLOGICAL WAR: A SCENARIO, which is unnerving.
Will post more as I learn more, including from Russian speaking sources🧵
shorturl.at/De4kW
Tom Knight -- e/🦀🌴 retweeted
Most of the anti-college sentiment today comes from people on the political right who don't share your values
This puts you in an awkward place but a powerful one.
The best world is a secret third world that cares about science (like the left) but ignores credentials (like the right). You are uniquely positioned to help build that world.
Tom Knight -- e/🦀🌴 retweeted
Should move BSL 3/4 labs to be fully robotic. These are one of the most dangerous places in the world for humans to work and we often put junior technicians in them and hope for no mistakes.
Russian authorities are responding to an incident in which a suspected dangerous infection was released as a result of a laboratory accident in Siberia and a researcher died. cnn.it/3Ubwb4F
Tom Knight -- e/🦀🌴 retweeted
you still have to do the hard stuff: to build an ai company in bio, you also need to build a bio company.
joehorsman.substack.com/p/he…
Tom Knight -- e/🦀🌴 retweeted
The Economist published an article yesterday calling for regulations to protect against mirror life. They propose restrictions on buying mirror chemicals, including extremely normal stuff like L-Glucose or D-amino acids. What is going on, guys? Where did this come from? I think the concerns about AI designing pandemics are overblown, but at least designing viruses is a real thing that we actually know how to do.
1. To be clear: we do not have any idea today how to make an autonomous cell from scratch, mirror or not. Making a mirror cell from scratch is much, much harder than making a non-mirror cell from scratch. I am a believer in strong AI acceleration but there is a lot of hard science and manufacturing that is missing here. People have ideas about how to do it, but they have ideas in the same way we have ideas about quantum computing: we will have lots of hints that it's around the corner before you see in on a shelf at walmart! Call me when we create an autonomous non-mirror cell in the lab from scratch, and then we can chat about sensible regulations.
2. Even if they existed, mirror bacteria wouldn't be able to escape the lab: they couldn't eat anything. Good luck being a left handed organism in a world where all the food is right handed. I know, a malicious person with access to mirror life could in principle give the mirror organism a way to digest non-mirror food, and then it could wreak havoc. But the converse is also possible: if a mirror bacterium were on the loose we could just give the normal non-mirror bacteria enzymes that would allow them to eat the mirror bacteria, and drive competition. This fact is generally ignored.
3. Finally, mirror life doomers pretend we have no defense against mirror life. This is also just nonsense. We have plenty of defenses. They are called mirror antibiotics. Mirror antibiotics would work against mirror life the exact same way non-mirror antibiotics work against non-mirror life. And there are way more options for mirror antibiotics than there are for non-mirror antibiotics, because we don’t have to worry about accidentally killing the human in the process! We actually even already have a ton of mirror antibiotics, since they are usually byproducts of antibiotic manufacturing. Fosfomycin, for example, is extremely simple, easy to manufacture in bulk, and industrial quantities of mirror fosfomycin are already produced today as a byproduct of fosfomycin synthesis. D amino acid analogs, like D-cycloserine, would also be very effective. If we can figure out how to make mirror life we can definitely figure out a bunch of universal mirror antibiotics to kill it.
Spare a thought for the poor scientists studying chiral biochemistry. Their requests to the models are all about to be denied as potentially dual risk. Meanwhile, want help finding a supplier for a toxin with nanogram LD-50? Sure, Claude will help you with that.
I agree mirror life escaping the lab would be bad and it deserves consideration, but can we stop pretending that it is a huge danger right around the corner that deserves front page regulatory attention, and instead focus on real shit like AI cyberattacks and the fact that the poles are still literally melting?
Tom Knight -- e/🦀🌴 retweeted
All hail @AnthropicAI & @DarioAmodei - you have been spared from horrific global calamity by Claude blocking me from compiling from the literature a list of zinc finger proteins which show poor expression in E.coli
Tom Knight -- e/🦀🌴 retweeted
All this renaming has me thinking Synthetic Biology could be more awesome.
From now on we do
ALT ULTRABIO
Tom Knight -- e/🦀🌴 retweeted
Not only models but now chemarxiv rejected my manuscript because it wasn’t “peer reviewed for harmful content” generated by AI. Peer reviewed for a preprint site? The bug is biting everyone and it’s turning into a pandemic.
Tom Knight -- e/🦀🌴 retweeted
Today, along with @Apheris_AI, we're introducing the Antibody Developability Consortium, with @AbbVie, @argenxglobal, @LundbeckUS, and @TakedaPharma as founding members.
The Consortium helps pharmaceutical and biotech companies predict manufacturability and developability risks earlier by building the field’s largest standardized antibody developability dataset.
Read the full press release here: hubs.la/Q04yyRk10
Tom Knight -- e/🦀🌴 retweeted
Astonishing how bad @AnthropicAI's biosafety filters remain, even after this long.
"Sometimes flag[s] biology-research-adjacent work" for obviously benign requests is pure marketing nonsense.
Absolutely amateurish for a company that claims to sell intelligence
Tom Knight -- e/🦀🌴 retweeted
You clearly are utterly and complexly clueless
The odds of current Claude models bricking themselves on innocuous biology queries is far closer to 100% than 0%
Charging customers for your rank incompetence is a terrible strategy
Tom Knight -- e/🦀🌴 retweeted
And Opus 5.5 won’t tell me the lengths of amino acid sequences I designed with *Claude*
I have the sequences already! Do you think not telling me the lengths will prevent human extinction?!??!????????
Tom Knight -- e/🦀🌴 retweeted
I wish we didn’t need these again, but here is the honest version of Anthropic’s biology announcement 🥳🥳🥳
(Caveat: I haven’t worked in bioinformatics for many years.)
The good: Anthropic ran ~950 Claude agents over a large biological sequence database. Claude searched, wrote code, compared sequences and genomic neighborhoods, and found an interesting pattern that apparently had not been noticed before: a known reverse transcriptase associated with another gene and a repetitive DNA array.
That is cool. Automating this kind of open-ended bioinformatics search at scale is useful, and Claude may have found a lead a human would have missed.
But: Claude did not do a biological experiment. It searched databases and analyzed data.
Humans then took the candidate into the wet lab. And the wet-lab result so far is modest: they showed that the repeat array produces short RNAs.
We still don’t know what the system does. No function, mechanism, phenotype, targeting, defense activity, or programmability has been demonstrated.
This is also where the CRISPR framing gets ahead of the result. Right now, “it has some features reminiscent of known programmable systems” is a hypothesis for what to investigate next, not a discovery that it behaves like CRISPR.
And there is a missing baseline: bioinformatics has had tools for finding unusual gene neighborhoods and candidate systems for years. The interesting comparison is 950 Claude agents vs. an expert using the best existing computational pipelines - not Claude vs. someone manually looking through 200,000 sequences.
So my honest announcement would be:
Claude autonomously found an interesting candidate for a previously uncharacterized biological system. A small human wet-lab experiment confirmed that part of the candidate is expressed. We don’t yet know what it does.
That is a good result.
But in a regular biology lab, this isn’t the finished paper. It is the result you show at lab meeting and say: “This looks interesting. Now we need to figure out what the hell it does.”
Maybe that next step leads to a major discovery. But that discovery hasn’t happened yet.
Claude has discovered a previously unknown enzyme system hidden in the DNA of bacteriophages. Beside the enzyme’s gene sits a long array of repeating DNA—a structure that looks somewhat similar to CRISPR.
We don’t yet understand what this system does, but only a handful of known systems share its features, and all of them are able to cut, copy, and paste DNA. Historically, the discovery of such programmable systems has helped revolutionize medicine. CRISPR, for instance, is now the foundation of genetic medicines. But it will take much more work to learn what this system does, and whether it can be put to similar use.
Read more: anthropic.com/news/claude-di…
