
It's not like we didn't choose the data and give the directives
Clinical trials of the first drugs designed with the help of artificial intelligence could commence this year, Google DeepMind CEO Demis Hassabis suggested Tuesday. Speaking on a panel at the World Economic Forum in Davos, Hassabis, who also runs DeepMind drug-discovery spin-off Isomorphic Labs, said he expected to have "some …
It's going to be one huge virtuous AI circle shirk.
At one end destroying the planet with its unquenchable thirst for energy and at the other keeping us all living forever so we can watch the real big bang - when the planet explodes because we've fucked it so hard.
Should be some spectacle.
You shouldn't conflate things like DeepMind with LLMs like ChatGPT.
DeepMind has already demonstrated its ability to find candidate molecules much faster than would ever be possible in the lab and elsewhere similar approaches have, for years, been used in pre-lab simulations based on our existing understanding, as much of this really is extremely advanced statistical modelling.
"candidate molecules"
i.e. modifying existing molecules to 'invent' 'new', patentable molecules, hopefully with no more side effects than the old ones had.
Efficiency isn't really important, *anything* can be sold as 'new and better' , with higher price and ridiculous profit, of course.
Once that is done, stop producing the 'obsolete' version as it produced less profit.
That's how profits are made: Minimum input, maximum profit.
"much of this really is extremely advanced statistical modelling." .... i.e. 'we have no idea'.
'AI' is just a lot cheaper than people in lab and therefore more profit. Even if >90% of the 'results' are pure BS, it's still cheaper than a lab.
You might want to read up on the reasons for the Nobel prize awards, but also the law: in most jurisdictions molecules themselves cannot be patented. The patents are almost invariably for the method to produce them. I know this routinely comes up, particularly with US companies, hoping to patent molecules they find around the world, but thus far largely unsuccessfully.
And it's not just be about being cheaper, it's about doing some tasks that could not be done in the lab otherwise within a reasonable time frame. For Google, in an analogy with the gold rush, this is more about selling picks and shovels, than gold nuggets.
Corporate dickhead makes yet another bullshit prediction for the benefit of compliant journalists.
Kissing the arse that feeds IT perhaps?
Oh, that reminds me, your RSS feeds are no longer linked anywhere, Are they going to be removed to promote advertising that we block anyway?
I am immune to "news" about "AI" now, so I did not read this particular article before commenting, I admit that.
This is what happens when a news site cries wolf too often and conflates entirely different concepts under the same buzzword. There are only so many hours in a day and I _need_ to filter.
So yeah now I have glanced at it I see that big companies will make shitloads of cash out of the general public. Big whoop.
If you read whole thing sideways it says that they created a random (or at least semi-random) protein generator, which checks if the generated protein is already patented.
.. and then they patent it. They have another "AI" writing patent applications, of course.
Hardly worth a Nobel, I say.
In the old times this would have been called "an expert system", not "AI".
I don't think that's quite correct.
AlphaFold is a deep-learning system1 which predicts a protein's tertiary (3d) structure from its amino acid sequence, with excellent accuracy. That's a really hard problem2, and AlphaFold a genuine breakthrough; hence the (worthy) Nobel prize.
The 3d structure of a protein accounts for its physiological effects; if you know the tertiary structure then, combined with many decades of accumulated physiological knowledge, you have a strong clue as to its likely effects. This means that, given a target physiological effect, you can randomly (or possibly better-than-randomly) generate amino acid sequences and throw away those whose tertiary structure is unlikely to achieve the target effect. This has the potential to greatly streamline the search for medically useful proteins.
(And yes, of course this will be monetised, and probably game the patent system, which is not hard as it's pretty broken.)
1Personally, I've long given up on getting exercised about what gets called "AI" these days. That bird has long flown.
2Back in the good ol' days before the advent of "AI", I was peripherally engaged with a group of phsycists and computational scientists who were developing software for predicting the secondary structure of RNA molecules from the base sequence. That's a way easier problem than protein tertiary structure prediction - but it was still really, really hard.
Hassabis has been hyping technologies that would deliver breakthroughs for decades now - he was originally a games developer promising artificial intelligence would deliver realistic interactions with computer generated characters in the game. That didn't happen.
Now we're being told "maybe soon" this latest set of damp future visions will come to pass... I'm not holding my breath.
I suggest you re-read the article, which is specifically about likely advances in AI this year, and not about the existing deep learning work that they got the Nobel Prize for.
Given the amount of money being thrown at this and other problems (and indeed, the wider field of AI), it would be extraordinary not to improve the state of the art for a lot of mathematical analysis tasks. On that front, AI as a whole is improving our technological toolkit. Continued funding at the current levels however depend on some significant breakthroughs that are - as Hassabis himself states - theoretical at best at present.
We shall see what comes next..
@Andy 73 "I suggest you re-read the article, which is specifically about likely advances in AI this year"
No the article was not. The only speculation of 'likely' happening this year was the opening 5 lines of the article about possible clinical trials.
"Clinical trials of the first drugs designed with the help of artificial intelligence could commence this year, Google DeepMind CEO Demis Hassabis suggested Tuesday.
Speaking on a panel at the World Economic Forum in Davos, Hassabis, who also runs DeepMind drug-discovery spin-off Isomorphic Labs, said he expected to have "some AI-designed drugs in clinical trials by the end of the year… That's the plan.""
You get a Nobel prize for doing something, or thinking something nobody else did. Nobel Prize winners are often people who also have other crazy ideas, but ones that turn out to have no basis in reality.
Hassabis is very smart, and has lead an effort which completed a very difficult task. Also, he's got what may be another crazy idea that doesn't work out.
By the way, I read a description once of Alan Turing: "There were some very smart people there, and I would often think -- 'I would have done that if I was smarter' --- but A.T. would come up with an idea, and I'd think --'I would never have thought of that' -- "
> Nobel Prize winners are often people who also have other crazy ideas, but ones that turn out to have no basis in reality.
Speaking as a mathematician and research scientist myself, I can confirm that pretty much every researcher (self included!) has crazy ideas that don't work out. It's part of the game. We usually don't hear about those people/ideas, though, as they're not famous.
> Hassabis is very smart, and has lead an effort which completed a very difficult task. Also, he's got what may be another crazy idea that doesn't work out.
It doesn't sound that crazy to me at all - the ability to actually implement that difficult task seems to me to open up genuine opportunities (see my earlier post). Whether it works out or not, we'll have to wait and see (probably a good few years, as clinical trials are generally1 a painfully slow business).
1Unless insane resources are thrown at them, as we saw with the Covid vaccines.
In that case, the bulk of medical advances, since forever, are useless (at least in the Western world, where this system is popularly known as "capitalism").
(Hint: although one can happen without the other, making money and helping ill people are not necessarily mutually exclusive.)
There's a lesson here for humans, as well. As you can see from posts on this very article, people (primarily Anonymous Cowards for some reason) have an assortment of bad takes, assuming that DeepMind is just a glorified chatbot and/or that nothing of value will arise from this new research tool. A rational, discerning person will look at the new evidence being presented and judge accordingly, rather than leaping to conclusions based on prior assumptions.