But more interesting is the 50% win rate stat that represents expert human performance in the paper.

That seems absurdly low, most employees don’t have a 50% success rate on self contained tasks that take ~1 day of work. That means at least one of a few things could be true:

1. The tasks aren’t defined in a way that makes real world sense

2. The tasks require iteration, which wasn’t tested, for real world success (as many tasks do)

I think while interesting and a very worthy research avenue, this paper is only the first in a still early area of understanding how AI will affect with the real world, and it’s hard to project well from this one paper.

https://cdn.openai.com/pdf/d5eb7428-c4e9-4a33-bd86-86dd4bcf1...

Tesla is uniquely risk tolerant for better or worse. You also don’t hear about people getting into accidents in a BMW on self driving because they don’t make the same claims and have tons of safeguards.

https://www.nature.com/articles/nrm3920

I'm not sure what you're implying here. Are you saying both types of structures are useful, but not as useful as the hype suggests, or that an X-Ray Crystal (XRC) and low confidence structures are both very useful with the XRC being marginally more so?

An XRC structure is great, but it's a very (very) long way from getting me to a drug. Observe the long history of fully crystalized proteins still lacking a good drug. Or this piece on the general failure of purely structure guided efforts in drug discovery for COVID (https://www.science.org/content/blog-post/virtual-screening-...). I think this tech will certainly be helpful, but for most problems I don't see it being better than a slightly-more-than-marginal gain in our ability to find medicines.

Edit: To clarify, if the current state of the field is "given a well understood structure, I often still can't find a good medicine without doing a ton of screening experiments" then it's hard to see how much this helps us. I can also see several ways in which a less than accurate structure could be very misleading.

FWIW I can see a few ways in which it could be very useful for hypothesis generation too, but we're still talking pretty early stage basic science work with lots of caveats.

Source: PhD Biochemist and CEO of a biotech.

Always remember, no decision is a decision. Usually that's the worst choice because almost any decision, even a wrong one, at least moves the ball in some direction, allowing you to gather more information. The only guarantee in this game is that stasis will kill you, so bias towards action. When in doubt, try to evaluate "most probable bad outcome" which is different from "worst possible outcome".

Good luck!

We've done work in this area and will be publishing some results later in the year.

Also shameless plug: I started a company to do just that, anchored to generating custom million-to-billion point datasets and using ML to interpret and design new experiments at scale.

"The theory goes that arm64’s fixed instruction length and relatively simple instructions make implementing extremely wide decoding and execution far more practical for Apple, compared with what Intel and AMD have to do in order to decode x86-64’s variable length, often complex compound instructions."

Not sure it's true, not an expert. But it doesn't sound wrong!

https://climate.nasa.gov/faq/33/which-is-a-bigger-methane-so...

Short of that, it's a luxury and a danger.

Self driving cars are a nice luxury, especially in city driving, not something that radically improves our world. You get to read your phone instead of paying attention, and the trade-off is someone might get killed. It's like treating a rash with a drug that could give you a heart attack. That's a far cry from, "with this technology something that took days and $$$$ now takes hours and $" as was the case with all the older examples you listed.

If self driving cars were more like airplanes, I'd have a little more faith. Tesla's marketing BS doesn't inspire me with lots of faith.

On black boxes: https://health.clevelandclinic.org/what-does-it-mean-if-my-m...

The problem is that in your scenario, many many more humans would be harmed. While animal models are flawed, there are no cell culture tools that even come close to recapitulating the system wide effects you get in a whole organism. Without animal tests, all the drugs that hurt a mouse (and there are very many that fail in mouse tests for this reason) would instead do it to humans.

You can ban animal research, but if you do, know that virtually none of the medicines you take would exist. It really is that stark.

1.) We don't know when to give the drug. Maybe giving it even earlier would help, but these trials don't tell us that. Answer: Run a new trial.

2.) We don't know how much drug to give. The drug was approved on the (bad, weak) evidence that in one high dose arm of one of 2 trials, there might have been an effect. Is that the right dose? Who knows! In the other trial, the high dose may have actually been worse. You can't titrate dosage in Alzheimers like you do in cancer where you can just watch how the tumor is shrinking. Answer: Run a new trial.

Giving this drug is not without downsides. You will have side-effects, including serious ones such as potentially brain swelling. Some people may be seriously injured or killed as a result of taking this drug. You have to make sure that the benefits outweigh that downside, and the trials show us a very dubious, weak effect.

The FDA should have said, "Good work, maybe there's an effect with this dosage, go run a new trial with the revised protocol." That's the right call.

Instead, they added, "Oh and you can sell the drug in the meantime and you don't have to tell us for 9 years."

How are you going to recruit patients for the trial? "We could give you the drug, but might give you a placebo." How many patients sign up for that instead of saying, "OR I could go out and buy the drug (which you claim totally works, and the FDA agrees!) independently."?

What if the trial fails in 9 years after you have tons of anecdotal reports (remember placebo shows an effect in past trials)? Now you have to take if off the market, imagine the loss of credibility that will entail for the FDA.

How about other drugs that reduced amyloid but showed no cognitive effect? (Eli Lilly's Solazenumab among many, many others) Should they get approved now too?

This is a mess for everyone and benefits Biogen. Everyone else loses, even the well-meaning patient advocates who created the political pressure for this decision.

Additional sources:

https://www.clinicaltrials.gov/ct2/show/NCT01900665

https://blogs.sciencemag.org/pipeline/archives/2019/12/06/th...

If competing experts are the question, note that 3 actual experts have resigned from what are coveted positions in protest. Nearly every part of the pharma industry (including the press, investors, other companies) who doesn’t stand to profit (I.e. not Biogen) has been up in arms saying this is an awful decision using words such as “horrifying”. There is no expert disagreement.

People can try to ret-con this by saying it’s like HIV, but note that viral load is a pretty good marker for disease morbidity in most viral infections. Amyloid is nothing like that as a validated marker for disease burden.

Useful sources:

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5797629/#s3titl...

https://endpts.com/what-does-a-clear-majority-of-the-biophar...