Please create an account to participate in the Slashdot moderation system

 



Forgot your password?
typodupeerror
×
AI Science

Former Google CEO Eric Schmidt Bets AI Will Shake Up Scientific Research 30

Eric Schmidt is funding a nonprofit that's focused on building an artificial intelligence-powered assistant for the laboratory, with the lofty goal of overhauling the scientific research process. From a report: The nonprofit, Future House, plans to develop AI tools that can analyze and summarize research papers as well as respond to scientific questions using large language models -- the same technology that supports popular AI chatbots. But Future House also intends to go a step further. The "AI scientist," as Future House refers to it, will one day be able to sift through thousands of scientific papers and independently compose hypotheses at greater speed and scale than humans, Chief Executive Officer Sam Rodriques said.

A growing number of businesses and investors are focusing on AI's potential applications in science, including uncovering new medicines and therapies. While Future House aims to make breakthroughs of its own, it believes the scientific process itself can be transformed by having AI generate a hypothesis, conduct experiments and reach conclusions -- even though some existing AI tools have been prone to errors and bias. Rodriques acknowledged the risks of AI being applied in science. "It's not just inaccuracy that you need to worry about," he said. There are also concerns that "people can use them to come up with weapons and things like that." Future House will "have an obligation" to make sure there's safeguards in place," he added.
This discussion has been archived. No new comments can be posted.

Former Google CEO Eric Schmidt Bets AI Will Shake Up Scientific Research

Comments Filter:
  • Shake "up?" (Score:5, Funny)

    by Narcocide ( 102829 ) on Friday November 03, 2023 @01:53PM (#63977532) Homepage

    No, I think you mean "shakedown."

  • by Alain Williams ( 2972 ) <addw@phcomp.co.uk> on Friday November 03, 2023 @01:54PM (#63977540) Homepage

    That would be interesting. Would anyone notice ?

    • In the case of Eric Schmidt, probably.... He's the one responsible for turning Google from the starry eyed public service company of the early years into the shithouse advertising and spy organization it has become. How could an AI do possibly worse?
  • AI will have to make many very great strides in order to do it with half the skill of a trained researcher.
  • This must be one of the more reasonable and realistic applications of AI that I came across since the hype started. Going through hundreds of papers is a very important but horribly time-consuming step of the research. Would be great if you can have an AI go through millions of them and then ask it for the answer to a question. Will not replace the need to actual do some reading yourself, but then the AI can point you to the most important papers or books. However, I can't imagine Elsevier, Springer or some
    • There are already too many bullshit papers flooding research databases. Combine "publish or perish" academic incentives with cheap tools that can boost ambitious, career-oriented (or just plain desperate) researchers & what we'll get is more of a tsunami of bullshit papers. Only recently has anyone started putting actual numbers on what we know is already happening to some degree, & it ain't pretty: https://arxiv.org/abs/2309.158... [arxiv.org]
    • This must be one of the more reasonable and realistic applications of AI that I came across since the hype started. Going through hundreds of papers is a very important but horribly time-consuming step of the research. Would be great if you can have an AI go through millions of them and then ask it for the answer to a question.

      But that would work only if you could then ask it how it came to give a particular answer. AFAIK, you can't ask LLMs how they came to give a particular response.

  • Scientists are already using AI to post bogus papers.
  • by backslashdot ( 95548 ) on Friday November 03, 2023 @02:13PM (#63977626)

    You'd be mad to do anything important with AI. AI can be used as an assistant to someone knowledge in the art but then so can a search engine and knowlege base.
    And yes of course i've used chatGPT, bard, midjourney etc. that's why I am saying AI sucks. They can improve AI over the next decade in certain specific areas such as factory automation and certain types of diagnostics but beyond that it's diminishing returns and whack-a-mole. If we are talking human-like AGI intelligence .. that's a minimum of 50 to 100 years away. Most likely well over 100 years away.

    • They are using it for more than that. In my mind I view it as a Python model learning app but they call it AI. They use AI to analyze breast exams, analyze MRI and look for tumors, and more. It displaces Radiologists. They started to displace them 10 years ago by outsourcing radiology analysis to India. Now AI is being used by less competent junior techs using a radiology AI app. It's creeping into many roles. Even in IT.
    • by gweihir ( 88907 )

      At this time? Very much so. With the current hype, the "ordinary person" interface of AI has gotten a lot better, but the quality of the answers has gotten worse. As to AGI, there is zero indication from what we currently have that it is even possible. It is not a question of computing power or training data either or we would at least have a glimmer of AGI today. Instead, we have absolutely nothing.

    • You'd be mad to do anything important with AI

      That's saying too much. AI has vastly improved things like handwriting recognition, for example.

  • Erich Schmidt is a serial philanderer and a genuine jerk. https://airmail.news/issues/20... [airmail.news]
    He was the one who said "Google isn't free, the cost is your information".
  • I'm not sure what part of the scientific method is compatible with inserting a black box complex enough that no one knows exactly what the processing path is.

    "You know what would help this two variable problem? About 5,000 more variables that are completely unknowable." - Einstein, probably

    • "About 5,000 more variables that are completely unknowable."

      They're completely knowable, it's just that there are a lot of them so they're hard to reason about.

  • Finally they'll use AI to fake the results data and graphs, so that's it's not as easily found out as today.

  • Seriously, why do people think that having money and some expertise in one area makes them experts in all others?

    One thing I can see happening is making plagiarism, fake results and low-quality research much harder to publish. That would be a good thing. As to the actual work of having insight and turning that into something useful, "AI" will contribute exactly zero. Also, a lot of applied researchers have had insights while dealing with more mundane stuff. I had the core idea for my PhD when thinking about

  • "Future House will "have an obligation" to make sure there's safeguards in place," he added.

    Nope. Won't work. It doesn't matter how hard they try to stop it, the cat is going to escape the bag. Maybe it's a data breach, maybe it's a new CEO who branches out licensing to questionable partners... but it'll get out. And the ratio of philanthropic uses to nefarious ones is at the very best 1:1.

    Not to be unreasonably fatalistic... but fatalism is appropriate here. It's coming. People will design new chemical

  • That the guys who think they hold the reins of power don't have a fucking clue what they're talking about.

  • AI is only as good as it's inputs, and right now science broadly is facing serious questions about quality [wikipedia.org]. What will help science is discipline and diligence to make sure published literature is legitimate; AI using current published literature is just going to be garbage in, garbage out.
  • Wow, what a savant to tell us all that AI might shake things up in (ALL sectors of) industry! What a genius that we all didn't deserve!

Avoid strange women and temporary variables.

Working...