Last month, Google announced “AI CO-Scientists“Ai the company is designed to help scientists create hypothesis and research plans. Google masced is the way to find new knowledge, but experts like – a briefly suddenly.
“This early tool, when interesting, it doesn’t work very well,” Beery Beery, compassionate vision researchers in mit, telling techcrroch. “I’m not sure that there is this type of hypotheses system provider from scientific community.”
Google is the latest giant giant to advance the understanding that AI will speed the scientific research on a day, especially in literary areas such as biomedikine. In the early essay this yearCEO Opeciai Sam Altman says “Ai SuperinTeligent Tool” can “speed the discovery and innovation quickly.” In addition, CEO AMOdei’s CEO AMODEI is boldly arguing that AI can help formulating medicine for general cancerSee rankings-.
But many researchers do not consider AI today to guide the scientific process. Applications like Google Ai CO CO-scientists look more hype than anything, says, which is not supported by empirical data.
For example, at Blog post Describing AI CO-scientists, Google said the tool had refers to the potential in the area such as leukemia abukemia drug drugs affecting the bone marrow. But the result is not expected that “no legitimate scientists (he) seriously,” said Fivia Dubyk, School Pathologue with Northwest-Tucson Health Center in Arizona.
“This can be used as a good starting point for researchers, but (…) shortage of detail does not worry and not tell us to trust,” said Dubyk. “Low information provided makes it hard to understand that this can be useful.”
The first of Google has been criticized by scientific communities to trumpsthrough the largest contempt for ai Ai who should be advised to produce the result.
In 2020, Google admit One of the trained AI systems to detect the breast tumors to obtain better results than human radiology. Researchers of Harvard and Stanford publish Rebuttal in the Journal Naturesaid the deficiency of ways and a detailed code about Google’s research “damaged (d) scientific values.”
Scientists have also tempted Google to glossing about AI Ai tools in scientific discipline such as materials. In 2023, the company said about 40 “new materials” already synthesis With the help of one of the AI systems, which are called Gnome. But, outer analysis Found not one of these materials, basically a new net.
“We will not know the power and restrictions such as Google ‘CO-scientists’ until they experience multiple discipline,” Konkey, assistant Institute of Technology, tells the technology technical, telling techcrete. “AI often performs in a good controlled environment but may fail when applied on a scale.”
Complex process
Part of the challenges to develop ai tools for helping found scientific is to transfer the number of confounding factors. AI may be easy to use in the required area of the large amounts, as a narrower from a large number of possible lists. But it is less clear whether AI can include types solving problems-boxes that lead to scientific villages.
“We’ve seen during historically the most important scientific progress, like the development of the most important MRNA vaccine, driven by intuitive and curse on the skepticism,” Khudishsh said. “AI, as it is now, it may not match the replication.”
However, researchers’ researchers on Sony Computer Laboratory in Japan, believes that the tool such as Google Co-Scientific Tools.
If you see the original value of ai that can apply for technical difficult tasks or tedious acts, as a new academic literature or formatting for the following conditions provided. But not many requests in the scientific community for AI scientists that produce hypothesis, they say – a lot of assignment of the intellectual researcher.
“For many scientists, I was included, producing hypotheses as the most fun part of the job,” said technrroch. “Why do I want to break my fun into your computer, then left just a generative job to make a generative art researcher why people do what they do, and we stop proposal to the product that opens the section.”
Borry noted that often the hardest step in the scientific process is planning and implementing the study and analysis of the verification or argument of hypothesis – which cannot reach the current AI system. AI cannot use physical tool to perform experiments, supposed, and often performs a worse issue for restricted data.
“Most science cannot be done with a significant component of a new scientific component, as Google is a real science processing, which is the context of the past and researchers, and similar resources, and the same resources.”
Ai risk
Technical earsis and ai risk – such as tending to Hallucinate – Also make scientists on alert to study a serious job.
Khudangkihsh Tools Kudukukhsh Kashudhsh can stop producing sound in scientific literature, not advanced forward.
Has been a problem. New research Meanworthy “Science Junk” is a Google Scholar flood, Google’s free search engine for the Literature scholarly.
“The point that made AI, if you are carefully careful, can flood the science field with lower or misleading study,” Korandukhshh’s studies. “A lot of peer-review processes have been a challenge in fields like a computer science, at the top conference has been exponentially exponentially on the post.”
Even good designons are well designed to be experienced by the wrong, saying. When he would like to provide an idea of an idea and literary synthesis, if he said he would not trust ai today to execute it.
“This is a variety of tools that exist, but that’s not a job I’ll pick up so far,” she said, adding her to take a problem Many AI systems are trained and in the amount of energy that is usedalso. “Although all the ethical issues (…) are solved, AI is now unlikely for me to apply for work on one way or another.”
