Future spaceA non -profit organization supported by Eric Schmidt, which aims to build a “AI scientist” in the next decade, began its first important product: a platform and an API with AI -related tools designed to support scientific work.
Many, many newly established businesses are struggling to develop AI research tools for the scientific field, some with huge amounts of VC funding behind them. Technological giants also seem refreshing in AI for science. Earlier this year, Google presented the “AI co-scientist“The company said it could help scientists in creating cases and experimental research plans.
CEOs of AI Openai and Anthropic companies have claimant That AI tools could massively accelerate scientific discovery, especially in medicine. But many researchers do not consider AI today is particularly useful for guiding the scientific process, largely because of its unreliability.
Futureshouse was released on Thursday four AI tools: Crow, Falcon, Owl and Phoenix. Crow can search for scientific literature and answer questions about it. Falcon can carry out deeper literature searches, including scientific databases. The owl is looking for previous work in a particular area. And Phoenix uses tools to help design chemistry experiments.
‘In contrast to others [AIs]Futureshouse has access to a huge high -quality open -ended body and specialized scientific tools, ” register Futureshouse in a blog post. “These [also] You have transparent reasoning and use a multi -stage process to examine each source in depth […] With the chain of them [AI]T together, on a scale, scientists can significantly accelerate the rate of scientific discovery. ”
But by saying, Futureshouse has not yet achieved a scientific discovery or making a new discovery with AI tools.
Part of the challenge for the development of an “AI scientist” is to predict an unspeakable number of confusion factors. AI can come useful in areas where widespread exploration is required, such as reducing a huge list of capabilities. But it is less clear if the AI is capable of the type of out -of -the -box problems leading to bonafide discoveries.
TechCrunch event
Berkeley, ca
|
June 5
Book now
The results of the AI systems designed for science so far have been mostly sluggish. In 2023, Google said about 40 new materials were synthetic with the help of one of his AIS, called Gnome. Yet an external analysis He did not find one of the materials, in fact, pure new.
AI’s technical shortcomings and risks, as its tendency to give up, also make scientists want to approve it for serious work. Even well -designed studies could end up being contaminated by AI’s bad behavior, which is struggling with high -precision work.
Indeed, Futureshouse recognizes that AI tools – in particular Phoenix – can make mistakes.
‘We release [this] Now in the spirit of rapid repetition, “the company writes in its blog post.” Please give comments as you use it. “
