𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ (@J33P4) header
𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ (@J33P4) profile picture

Specialising in AI, biology and everything else interesting.

UK
Joined March 2021
Photo by 𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ (@J33P4)Photo by 𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ (@J33P4)Photo by 𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ (@J33P4)Photo by 𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ (@J33P4)Photo by 𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ (@J33P4)Photo by 𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ (@J33P4)Photo by 𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ (@J33P4)Photo by 𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ (@J33P4)Photo by 𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ (@J33P4)Photo by 𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ (@J33P4)
𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ retweeted
🚨There is 0% this is possible. If you are developing drugs πŸ”Ίdementia or diseases with long-term follow-up, πŸ”Ί 🚨You can’t have a 12-week study and say you cured dementia. A participant has to be followed at least 12 months to 24 months. Your trial at Phase 2 and 3 is going to be a minimum of 24 months; will say Phase 1 12 months (that’s 5 years in trials) Fastest 6, 7-8 years with some kind of breakthrough designation. No one’s fault that’s the biology. == Ok, let’s say discovery and even synthesis of a compound takes 1 day (takes weeks to months) I’ll even say animal studies and IND take 12 months. So with human trials even if everything goes perfectly. 🚨 So it’s 6 or 7-8 years to approval even with some kind of breakthrough designation. Also, they have zero drug candidates. A normal thing to say after this is β€œwe’re seeing progress in the lab”, but they don’t have anything. It’s not being negative. It’s just not possible to assess a long-term outcome in a short time frame.
Dario Amodei says cancer, Alzheimer’s, heart disease, and mental illness will be relics of the past in less than a decade, and that "we can create an age of miracles" β€œYou know, if today we look at things like cancer, Alzheimer’s, heart disease, mental illness, these will be relics of the past in, I think, less than a decade." "And I believe that the company which gets the risks right will also be the company that has the license to operate that society trusts and society believes can build the benefits in the right way." "I think if we do this right, we can create an age of miracles where humanity is able to solve problems that have eluded us for millennia." "We have to make sure that all the risks, all the ways that things can go wrong with this technology, we very carefully address those risks as we bring the benefits, and as we have a viable business at the same time.”
259
79
25
774
115,658
𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ retweeted
Pretty meaningless, the results were in the training data
JUST IN: OpenAI says its AI has solved the Navier–Stokes Millennium Prize problem
28
42
2
1,778
61,043
𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ retweeted
British Airways cancelled a flight, rep said BA at fault, compensation due, just complete e- form. I did. Claim then refused, BA declined to answer specific questions. Told to stop bothering them, appeal to CAA if I must. Just won the case. BA customer service is so poor
63
231
5
4,225
136,143
𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ retweeted
37
176
18
2,414
124,933
𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ retweeted
The accusation that I am antisemitic is appalling and fundamentally dishonest. Criticizing the actions of the Israeli prime minister, a military technology contract, or the executives who supply it is not the same as criticizing Jewish people. This critical and necessary dialogue is then dishonestly framed as being anti-Israel.Β To be clear, my views come from my own political convictions and should never be interpreted as hostility toward Jewish people, for whom I have deep love and respect.Β  Everything I know about acting, activism, and humanism has been profoundly shaped by the Jewish friends, colleagues, and loved ones who have been integral and family throughout every point of my life. This merger has real consequences for real people, and for the entire country. Scrutinizing the Ellisons, including Oracle’s business built on data, surveillance technology and government contracts, and the serious threat to editorial freedom and the loss of a livelihood for thousands of families, is fair and necessary. The $111 billion deal would hand one family control over CNN, HBO and Warner Bros., backed in part by foreign money whose influence on editorial decisions has never been fully explained to the public.Β  Lawmakers on both sides of the aisle have called for serious national security review, and regulators still haven’t given the public a real answer. Until they do, the merger shouldn’t move forward.Β Now is the time to dive boldly into all these issues, not step back or concede.
19,479
156,559
6,446
858,029
31,898,972
𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ retweeted
Sorry, but the only way to eliminate all diseases in 5 years is to eliminate all biological life itself. The fastest time you can approve a moderately-novel drug if you have a bulls eye target and maximally multiparameter optimized molecule or a biologic is 4.5 years.
Replying to @DarioAmodei
2/2 Second, on the messaging around AI.Β  I do not agree that my messaging has been disproportionately negative.Β  In fact it has been about equally balanced between risks and benefits: I’ve written one major essay about each, and even in interviews where I discuss the risks, I make sure to frequently mention the incredible benefits as well as proposing possible solutions to the risks (short clips from my interviews that end up on social media tend to be disproportionately negative, as that gets clicks).Β  In fact, I wrote Machines of Loving Grace because I didn’t feel the AI industry was painting an inspiring enough picture of how the technology could radically transform the world for the better.Β  The bulk of the essay is devoted to refuting skepticism of AI’s potential in health and biology, and showing why I think it will actually be possible to cure most human disease in ~5-10 years, as crazy as it may sound to ordinary people and frankly to biologists as well (I used to be one!).Β  And, if you read my most recent essay (Policy on the AI Exponential), I discuss concrete proposals for how to streamline the FDA process to make sure the deluge of AI-accelerated drugs isn’t slowed down by the regulatory process.Β  I feel the urgency here: I lost my father to Hepatitis C only a few years before the development of direct-acting antivirals (sofosbuvir), which cure 95% of patients and probably would have cured him. I do agree that the public has a negative view of AI (and that this is a big problem), but I don’t think it is primarily caused by me or any other AI leader warning about AI’s risks.Β  I think it is fundamentally a crisis of trust.Β  I think that ordinary people don’t trust companies, governments, or the tech industry and always suspect that we are cooking up some new way to screw them over.Β  The causes of this go back decades and AI is just the latest iteration of it.Β  I don’t think that a glitzy marketing campaign with a positive spin (which some have advocated that Anthropic do) is the way to win back that trust β€” at this point, saying that AI will cure cancer is more a cliche than it is inspiring, and most people think it is deceptive.Β  The thing that will work is *actually curing cancer*.Β  I think by far the most accurate criticism of AI companies including Anthropic is that we haven’t yet delivered on our big promises to benefit the world.Β  That is totally on us, and I think it’s the criticism you should be making, instead of all this stuff about messaging and marketing. We are however doing our best to fix this: Anthropic is ramping up its efforts very quickly in biology and medicine, and we hope to have incredible results in the coming years and some early glimmers in the coming months.Β  When we’ve actually accomplished something real, the whole world will hear about it, as loudly as possible, you have my word on that.Β  But until then I don’t want to make empty promises, and in the meantime I feel compelled to speak honestly about the very real risks of AI and how to address them.Β  Honesty is the right thing on the merits, and in terms of public credibility and trust it is no worse than, and may in fact be better than, an approach that ignores or distracts from risks which people instinctively understand are real.
14
17
8
143
18,510
𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ retweeted
Replying to @mattdeitke
Don't you think that this post is dishonest?
2
1
1
15
1,042
𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ retweeted
Only about a quarter of human diseases have an approved therapy. By some estimates it's a few percent. Most of those treatments slow a disease rather than stop it. Closing that gap is what the AI-cures-everything story promises. Build a system smart enough and the cures hidden in what we already know will fall out. I have worked at the intersection of machine learning and biology for three decades, and I believe #AI will eventually transform human health. The capabilities arriving now are extraordinary. But the promise rests on an assumption that is simply false: that we already understand human biology well enough for a clever enough reasoner to find the answers in it. We don't. More than 90% of drugs entering clinical trials fail, a number that has barely moved in decades. In the large majority of those failures the molecule was engineered just fine. The mechanism it targeted was wrong. We are doing a pretty good job at manufacturing keys, but they are generally for the wrong locks. And because nobody wants to fail in the clinic, the industry has retreated to the locks it already trusts: 38 targets now have more than 50 programs against each of them, while the number of novel targets advanced per year fell from roughly 100 in 2015 to about 30 in 2024. AI will not reason its way past this. Biology wasn't engineered. It is the product of billions of years of messy, stochastic evolution, and the variation that produced is too vast and too idiosyncratic to work out in the abstract. You have to measure it. Aimed at a biology this thinly sampled, AI will mostly help us generate failures faster. I founded @insitro because getting to the right locks requires a different kind of system. We generate multimodal human and cellular data at scale, use machine learning to find causal drivers of disease, and test those hypotheses experimentally. Virtual Humanβ„’ is built for causal discovery; TherMLβ„’ turns what it finds into the right therapeutic intervention. It is working: first-in-class programs internally and with partners, three #ALS targets that Virtual Humanβ„’ identified and @bmsnews nominated, and additional collaborations with @EliLillyandCo and @GileadSciences . Today we are launching Deep Phenotype: Scaled Biology, Deep Causality. Issue one is "Drug Discovery Has No Magic Wands," the first half of a two-part essay on the magical thinking currently running through our field and what I think it will actually take. After that you will hear from insitro's own scientists and engineers, people who work across computation and experiment because the problem requires both. Getting this right is hard, and we do not have all of it worked out. I hope you will follow along and think it through with us. DeepPhenotype.Substack.com
27
84
25
409
82,468
𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ retweeted
Wasserman's Filmmaker Suite now includes 8 total Apps/Mcps. Link to all of themπŸ‘‡. Free + Open Source. New - Script Break, Corkboard, Storyboard Reference Studio. Updated - Blockout. Also avail - Master Canvas, Motion Previs Studio, Unofficial Davinci MCP
11
14
2
71
3,061
𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ retweeted
Wanna know why Anthropic hates Opensource AI? Kimi K3 being free and available to download made their $1 Trillion valuation make no sense
Kimi K3 is only a couple of points below Fable 5 and GPT 5.6 Sol Max on the Intelligence Index Fable 5 wasn't even available for majority of time this model was being trained so they cannot scream "distillation" Opensource AI is now Frontier Intelligence.
114
344
17
4,122
135,523
𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ retweeted
PREDICTION after kimi k3 successful launch and mogs opus 4.8, Dario Amodei is gonna write an essay explaining how open-weight Chinese AI threatens democracy, national security, and the future of humanity
323
1,138
138
16,189
483,504
𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ retweeted
Repost if you think Gianni Infantino is the worst FIFA president ever.
1,974
40,364
687
85,239
1,003,447
𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ retweeted
Sigh. There’s no gatekeeping. Doctors wouldn’t do these scans for ourselves even if you paid us for it. We have been around the block, and know what happens when you get told β€œcarcinoma cannot be ruled out” over nothing. The knowledge is worse than useless.
Replying to @jakehalloran1
Twitter posters are smarter than basically every doctor they've visited but those doctors know that they're smarter than basically every patient. That basic fact explains a lot of the repetitive doctor discourse, because doctors gatekeep medical treatments
16
5
2
134
13,041
𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ retweeted
You’re not getting β€œmore information.” You’re getting something that tells an otherwise healthy and happy person that β€œcarcinoma cannot be ruled out.” Now you’ve committed them to several more scans, and then to potential biopsies and surgeries. And the expenses and complications of all that.
I just accepted that doctors are bad at math it is embarrassing that the field responsible for making life or death decisions just collectively agreed on the obviously false, grade-school disprovable statement that more information can be bad to a robust decision making process
46
4
6
187
19,982
𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ retweeted
πŸ™„πŸ™„You clearly don't work in medicine. Once a radiologist reports a significant finding, It's documented. At this point, the physician is legally and ethically obligated to work it up. Ignoring it is malpractice. That's why the prevailing feeling is that unnecessary imaging creates this trap. οΏΌοΏΌοΏΌIt's a finding that you can't unsee and just leave alone. The workup isn't free. Biopsies,contrast, radiation all of these things have real complication rates. Would you want some extra radiation after having a test that you probably didn't need? Or better yet, a large needle driven through one of your organs. How does that sound?
3
1
1
7
686
𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ retweeted
I used to wonder if Med Twitter existed whether Theranos would have still happened. Now I know the answer.
15
9
1
231
21,918
𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ retweeted
I use ultrasound many times every single day in the ICU, but there’s a lot of unreasonable hype about this β€œwhole body imaging” ultrasound. I’m very skeptical of the claims being made and I’ll explain why: First some fundamental limitations of US: Ultrasound doesn’t penetrate bone and doesn’t pass through air very well. This makes imaging the brain and lung parenchyma essentially impossible with US. Bowel gas also frequently makes it difficult to visualize abdominal structures like stomach, small bowel, colon. It can’t see into bone either which can limit its utility for musculoskeletal imaging. For this reason ultrasound isn’t really amenable to a β€œwhole body scan.” It’s hard to see how this could replace other modalities (MRI, CT) if it can’t visualize so much of the body. Ultrasound exams are often dynamic. If you’ve had one you may have been asked to roll or move to visualize certain structures. This requires a skilled operator. Immersing the patient in a tank for a 1 minute scan is a cute shortcut. It’s technically easier but it probably won’t be able get optimal images, further limiting interpretation. Also are they exchanging the water in the bath each time? Much of the time spend on imaging is actually cleaning the scanner between patients. Unclear how you can quickly disinfect a liquid scanner. The theoretical resolution of ultrasound is very high, but that isn’t quite the same thing as being able to identify structures. There are lots of artifacts and limitations to ultrasound. For some organs (thyroid, kidney, liver) it’s great. For others it may be less so (pancreas, colon, stomach, etc). Hard to see this replacing existing methods, especially if the concern is cancer screening. Ultrasound isn’t really one modality. There are a lot of different techniques (B mode, M mode, 3d modes, different types of Doppler, etc). Unclear how many of these this gizmo can do. Adding these capabilities may make the scan more capable but will also add to the scan time if it has to switch modes. There’s no free lunch. There’s always a tradeoff between scan quality and time. AI is great at *certain* narrow medical image interpretation tasks. But there isn’t a massive training set of data for this β€œnew” modality. I wouldn’t expect AI to be very good at reading these scans until they’ve accumulated millions. That means they are still paying human radiologists to interpret for the foreseeable future. Everything in medicine is based on evidence. Proving that lung cancer screening saves lives took a decade. Where are the studies for this? So far just hype. More concretely, without evidence insurance won’t pay. Finally, Who is this technique for? Yes it avoids ionizing radiation but so does MRI. Yes it’s quick, but so is a CT scan. The scan may be quick but the interpretation may be slow (It’s still dependent on human radiologists) and the machine may require time to clean. It can’t image lots of body parts so it’s hard to see how it replaces β€œwhole body MRI scanners.” I’m sure there are tech/wellness bros who are excited but pay out of pocket for low quality wuick partial body scans but wider adoption depends on more than hype.
Announcing a new division of Midjourney called "Midjourney Medical"
157
85
34
675
145,708
𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ retweeted
Two things are true: (1) Anthropic (or parts of it) are absolutely and sincerely worried about the misuse of Mythos-class models & have put in excessive safeguards until they are confident it will not be misused (2) They have not succeeded in explaining/convincing people of this
91
44
10
912
71,864
𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ retweeted
New Fable benchmarks from @claudeai
9
28
8
288
38,331
𝗝 𝟯 𝟯 𝗣 𝟰 | π—·πŸ―πŸ―π—½πŸ°.π—²π˜π—΅ retweeted
Hot take: They're not subsidized their margins are insane. They are just absolutely raping api customers. Anyone who has used deepseek or hosted anything and done the math on hardware/power costs knows this
Subscription plans are massively subsidized. And by massively, I mean absurdly: Claude Max 20x: $200/month, with usage reportedly worth around $8,000 ChatGPT Pro 20x: $200/month, with usage reportedly worth around $14,000
121
136
24
4,342
504,358