Scientists spent 10 years on a superbug mystery - Google's AI solved it in 48 hours

Shawn Knight

Shawn Knight

Posts: 15,581   +197
Staff member
Cutting corners: Researchers at Imperial College London say an artificial intelligence-based science tool created by Google needed just 48 hours to solve a problem that took them roughly a decade to answer and verify on their own. The tool in question is called "co-scientist" and the problem they presented it with was straightforward enough: why are some superbugs resistant to antibiotics?

Professor José R Penadés told the BBC that Google's tool reached the same hypothesis that his team had – that superbugs can create a tail that allows them to move between species. In simpler terms, one can think of it as a master key that enables the bug to move from home to home.

Penadés asserts that his team's research was unique and that the results hadn't been published anywhere online for the AI to find. What's more, he even reached out to Google to ask if they had access to his computer. Google assured him they did not.

Arguably even more remarkable is the fact that the AI provided four additional hypotheses. According to Penadés, all of them made sense. The team had not even considered one of the solutions, and is now investigating it further.

Co-scientist is a multi-agent AI system built using Gemini 2.0. According to Google, it serves as a "virtual scientific collaborator" that can help generate novel hypotheses and research proposals, and speed up biomedical and scientific discoveries. Research organizations interested in co-scientist can apply to participate in a trusted tester program.

AI has been a topic of debate for years. Among other concerns, critics warn that it could impact jobs and put scientists like Penadés out of work. The lead researcher told the BBC that he understands the fear, but believes that having an extremely powerful tool outweighs the negatives.

Penadés is now sold on artificial intelligence. "This will change science, definitely, he said, adding that he believes he is witnessing something spectacular. "It's like you have the opportunity to be playing a big match - I feel like I'm finally playing a Champions League match with this thing."

Image credit: CDC

Permalink to story:

Scientists spent 10 years on a superbug mystery - Google's AI solved it in 48 hours

 
Squid Surprise said:
Once again, for all those people saying we are wasting our money on AI… this is the kind of thing it is useful for…
Click to expand...
No one said "we are wasting money on A.I" people are saying the extravagant amounts of money is a waste, as the criticism comes from people seeing all this outrageous spending put mostly in competition against other companies instead of mostly going into making A.I truly useful at the pace it requires instead of for ing it down infringing on people's lives.
 
  • Like
Reactions: allis0
Where this will be good is turning highly dangerous pathogens into rather benign ones, this is already a strategy we use, same for cancer. Ie killing such stuff can be fraught with problems, eg side effects as attacks normal cells, tissues, and even deadly battle harden strain that can pass it's defense to other pathogens.
So AI can run billions of evolutions to find a mutation that is more benign and dominate other variants.

Lots of pathogens get there naturally no good if host dies and can't migrant - some even pass on a host of benefits and become symbiotic
 
I am not sure if it’s something that’s impressive. Essentially the LLM has been trained with massive amounts of data, which may include what these scientists are working on. So yeah, the data’s there, so the AI “figured” it out sooner is expected. It’s like you can spend 10 years to compile a list stored on excel, but only needs a fraction of the time to use control+ F to find it, or some v-lookup to find something quite similar.
 
Watzupken said:
I am not sure if it’s something that’s impressive. Essentially the LLM has been trained with massive amounts of data, which may include what these scientists are working on. So yeah, the data’s there, so the AI “figured” it out sooner is expected. It’s like you can spend 10 years to compile a list stored on excel, but only needs a fraction of the time to use control+ F to find it, or some v-lookup to find something quite similar.
Click to expand...
You obviously haven't read the article. Perhaps you should use your own advice and try Ctrl+F and search for "results hadn't been published anywhere" on this page. So yeah, kind of hard for the AI to simply search for and find information that didn't exist anywhere.
 
  • Like
Reactions: RancherBill and Squid Surprise
The Talking Tech said:
No one said "we are wasting money on A.I" people are saying the extravagant amounts of money is a waste, as the criticism comes from people seeing all this outrageous spending put mostly in competition against other companies instead of mostly going into making A.I truly useful at the pace it requires instead of for ing it down infringing on people's lives.
Click to expand...
Apparently you haven't been reading the comments on this site over the past several months...
 
Kashim said:
You obviously haven't read the article. Perhaps you should use your own advice and try Ctrl+F and search for "results hadn't been published anywhere" on this page. So yeah, kind of hard for the AI to simply search for and find information that didn't exist anywhere.
Click to expand...

This article ommits key information that contradicts the sensationalist headline. From the New Scientist article:
However, the team did publish a paper in 2023 – which was fed to the system – about how this family of mobile genetic elements “steals bacteriophage tails to spread in nature”. At the time, the researchers thought the elements were limited to acquiring tails from phages infecting the same cell. Only later did they discover the elements can pick up tails floating around outside cells, too.

So one explanation for how the AI co-scientist came up with the right answer is that it missed the apparent limitation that stopped the humans getting it.

What is clear is that it was fed everything it needed to find the answer, rather than coming up with an entirely new idea. “Everything was already published, but in different bits,” says Penadés. “The system was able to put everything together.”
 
  • Like
Reactions: allis0 and docbillnet
Very impressive. AI is also doing wonders in material science, being able to sort through millions of potential candidates for a certain set of metrics in mere weeks and reduce the list to a few dozen candidates. Traditional quantum chemistry approaches take years and years. They can now focus of jst simulating the small list of candidates and trying to manufacture them.
 
  • Like
Reactions: BadThad
AlexIL said:
Ao now they can replace scientists with AI
Click to expand...

It takes a scientist to propose an idea to a machine and to validate it's results. In science, many things are calculatable, but that does not make them accurate. Mother Nature has a way doing what it wants to do rather than what you think it will do.
 
daffy duck said:
Very impressive. AI is also doing wonders in material science, being able to sort through millions of potential candidates for a certain set of metrics in mere weeks and reduce the list to a few dozen candidates. Traditional quantum chemistry approaches take years and years. They can now focus of jst simulating the small list of candidates and trying to manufacture them.
Click to expand...

And I see the possibilities in chemistry. We've used various proprietary software over the decades to help make predictions on complex mixtures - but the process was cumbersome due to the amount of input data required. I suspect an AI version could reduce the amount of input data by using common chemistry and the general properties of the components. If AI understands the end goal, it could bring a formulator closer to a final product with fewer experimental steps.
 
The AI did not do ten years qorth of research. Even with the actual research results not available it had access to all the biology advances and all indirect references to the research, and all inspiring research. Had you just given AI ten year old data it eould have come no closer to hypothesising the correct solution than any graduate student from ten years ago interested in the subject.

These articles which mislead the public on AI abilities will result in research defunding and stalled progress.
 
Watzupken said:
I am not sure if it’s something that’s impressive. Essentially the LLM has been trained with massive amounts of data, which may include what these scientists are working on. So yeah, the data’s there, so the AI “figured” it out sooner is expected. It’s like you can spend 10 years to compile a list stored on excel, but only needs a fraction of the time to use control+ F to find it, or some v-lookup to find something quite similar.
Click to expand...
Did you even read the article? It specifically stated that nothing from this was published at all. The AI generated a primary and 3 other plausible solutions to a very complex research problem that required 10 years to arrive at the same primary conclusion.
 
  • Like
Reactions: Squid Surprise
*This* is where AI shines: Taking large complex datasets and finding all the possible ways you can get to a given result. It's *really* good at finding ways complex systems interact, something current models struggle with.
 
It would make a more convincing argument for AI if it was used to find solutions to problems we haven't yet solved yet.
 
You must log in or register to reply here.

Similar threads

midian182
AI adoption is on the rise, with 3 in 5 Americans saying it improves their lives
Replies
8
Views
113
Sir Sparkles
Sir Sparkles
Shawn Knight
The New York Times approves AI tools to assist journalists
Replies
1
Views
49
Kirby1
K
midian182
Lenovo's ThinkBook Flip AI laptop features a towering OLED screen that folds in half
Replies
4
Views
72
Megalomaniac
Megalomaniac

Latest posts

Back