While I’m still a bit leery of the art, I think my real problem with most of the stuff we’re calling “AI” is the context in which it exists, same as with other forms of automation. There’s also a reasonable concern about the energy requirements of this technology, but today we’re going to look at one of the was in which it could be a huge benefit to all of us.
The development of computing machines in the 20th century made a lot of things possible, including new avenues and scales of research. The climate models that have, despite what you may have heard, done a surprisingly good job at projecting future warming, needed the capabilities of computers. Like its predecessors, this new AI technology opens up new approaches to research that were previously off the table.
An artificial intelligence system enables robots to conduct autonomous scientific experiments—as many as 10,000 per day—potentially driving a drastic leap forward in the pace of discovery in areas from medicine to agriculture to environmental science.
Reported today in Nature Microbiology, the team was led by a professor now at the University of Michigan.
That artificial intelligence platform, dubbed BacterAI, mapped the metabolism of two microbes associated with oral health—with no baseline information to start with. Bacteria consume some combination of the 20 amino acids needed to support life, but each species requires specific nutrients to grow. The U-M team wanted to know what amino acids are needed by the beneficial microbes in our mouths so they can promote their growth.
“We know almost nothing about most of the bacteria that influence our health. Understanding how bacteria grow is the first step toward reengineering our microbiome,” said Paul Jensen, U-M assistant professor of biomedical engineering who was at the University of Illinois when the project started.
Figuring out the combination of amino acids that bacteria like is tricky, however. Those 20 amino acids yield more than a million possible combinations, just based on whether each amino acid is present or not. Yet BacterAI was able to discover the amino acid requirements for the growth of both Streptococcus gordonii and Streptococcus sanguinis.
To find the right formula for each species, BacterAI tested hundreds of combinations of amino acids per day, honing its focus and changing combinations each morning based on the previous day’s results. Within nine days, it was producing accurate predictions 90% of the time.
Unlike conventional approaches that feed labeled data sets into a machine-learning model, BacterAI creates its own data set through a series of experiments. By analyzing the results of previous trials, it comes up with predictions of what new experiments might give it the most information. As a result, it figured out most of the rules for feeding bacteria with fewer than 4,000 experiments.
“When a child learns to walk, they don’t just watch adults walk and then say ‘Ok, I got it,’ stand up, and start walking. They fumble around and do some trial and error first,” Jensen said.
“We wanted our AI agent to take steps and fall down, to come up with its own ideas and make mistakes. Every day, it gets a little better, a little smarter.”
Little to no research has been conducted on roughly 90% of bacteria, and the amount of time and resources needed to learn even basic scientific information about them using conventional methods is daunting. Automated experimentation can drastically speed up these discoveries. The team ran up to 10,000 experiments in a single day.
This is the kind of thing that could, at least in theory, lead to the rapid development of new antibiotics – something we definitely need. I think the odds are good that even without using robots to do experiments, this technology could accelerate a great many fields of research. I’m not looking for some kind of magical tech “solution” to climate change or pollution, but there’s a real possibility that advances due to this tech could enable us to survive conditions in the future, which would destroy us in the present.
Of course, that all depends on who gets to decide what kind of research is prioritized, and what is sidelined. Our technology may give us the ability to do marvelous things, but it will never fix the injustice, inequality, and suffering that are a deliberate outcome of our current political and economic system. Technology won’t ever remove the need for revolution, all it will do is change the landscape in which we fight.
This is a fun proof-of-concept test of AI research, but it doesn’t seem terribly useful – figuring out which amino acids different species like, when in reality bacteria probably consume and metabolize hundreds of thousands of unique organic molecules in natural settings, isn’t that interesting. It’s a pretty major leap away from learning anything about functional differences between species, and that takes directed, intentional research, which AI hasn’t been able to do yet.
It seems to me that the research community is at least as much apprehensive of AI as they are hopeful. There is a fear that it could fuel an increase in junk science – such as AI-written abstracts or methodology that is more misleading than accurate – while the benefits are still fairly uncertain. I can see a few potential applications in accelerating drug discovery, but your example – developing new antibiotics – is almost as much an economic problem as it is a scientific one: new antibiotics aren’t profitable, so all of the major pharmaceutical companies have completely abandoned the market. AI won’t fix that.
Fair points.
It seems like a lot of people are worried about AI being used to pump out so much useless garbage that it’s hard to even find real work. Given the way the incentives of our society are set up, it’s a reasonable fear.
But I still think that there’s a lot of potential, well beyond this particular study, for real advancement. As to the profitability of new antibiotics, I guess I’m not convinced by that argument. There seems to be zero question that the demand for them will increase as drug resistance spreads.
The linked article does not make it clear what it means to do “hundreds of combinations of amino acids per day” (though a photo seems to show mechanical hands placing a test tube in a tray).
If this approach involves physical handling, combining, and assessments, it seems that still requires a human to prepare the equipment, provide the samples to the machinery, wash up afterwards, etc. So scientists can take some comfort in knowing they will still be needed to serve as washers and carriers in the brave new laboratories of the near future…
@2,
The economics of antibiotic development is a real problem, and it’s unique to antibiotics. The problem stems from the fact that once we have a new drug, we don’t want to use it – we want to stick it on a shelf behind “Emergency Use Only” glass and save it for highly drug resistant infections. We don’t want to develop a brand new drug and immediately start using it for everything, causing resistance to the new drug to become widespread.
That’s not something that regular supply-and-demand fixes. That’s FDA policy (and EMA, and other major drug regulators). We desperately need new drugs, but once we have them, we want to use them sparingly.
If new antibiotics could be profitable, big pharma would be buying up the patents. But they’re not. They’ve all closed down their own research labs (my partner was on Novartis’s last remaining antibiotics research team), All the research and clinical trials are done at small pharma companies, often running on public grants because investors know they’re not profitable. Even some companies that have managed to get a new antibiotic approved still folded after the approval (e.g., Achaogen).
@invivoMark
That makes sense. Thanks for taking the time to explain!