{"id":462222,"date":"2026-06-05T10:21:10","date_gmt":"2026-06-05T10:21:10","guid":{"rendered":"https:\/\/www.newsbeep.com\/nz\/462222\/"},"modified":"2026-06-05T10:21:10","modified_gmt":"2026-06-05T10:21:10","slug":"could-ai-research-assistants-speed-up-scientific-discovery-research","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/nz\/462222\/","title":{"rendered":"Could AI research assistants speed up scientific discovery? | Research"},"content":{"rendered":"<p class=\"picture\"><img fetchpriority=\"high\" decoding=\"async\" alt=\"AI assistance\" src=\"https:\/\/www.newsbeep.com\/nz\/wp-content\/uploads\/2026\/06\/548794_gettyimages2248049740_426350_crop_crop.jpg\"  loading=\"eager\" class=\"lazyloaded\" width=\"6000\" height=\"4000\"\/><\/p>\n<p>Artificial intelligence (AI) research labs Google DeepMind and Futurehouse have now released AI research assistants that can generate scientific hypotheses, design experiments and analyse data. The researchers behind these systems think that they could help accelerate scientific discovery, for example, by identifying clinically approved drugs to treat unrelated conditions or diseases.<\/p>\n<p>AI tools have so far helped scientists analyse large amounts of data, <a title=\"How AI protein structure prediction and design won the Nobel prize\" href=\"https:\/\/www.chemistryworld.com\/features\/how-ai-protein-structure-prediction-and-design-won-the-nobel-prize\/4020354.article\" rel=\"nofollow noopener\" target=\"_blank\">predict the structure of proteins<\/a> or even <a title=\"As AI-designed drug looks to pass final hurdle will this tech change drug discovery forever?\" href=\"https:\/\/www.chemistryworld.com\/news\/as-ai-designed-drug-looks-to-pass-final-hurdle-will-this-tech-change-drug-discovery-forever\/4021894.article\" rel=\"nofollow noopener\" target=\"_blank\">design small molecule drugs that are starting to enter clinical trials<\/a>. \u2018[But] how do we teach AI systems to think in the manner that scientists do?\u2019 says <a title=\"Vivek Natarajan | Google\" href=\"https:\/\/research.google\/people\/106654\/\" rel=\"nofollow noopener\" target=\"_blank\">Vivek Natarajan<\/a> at Google DeepMind. Current interactive AI tools \u2013 such as Chat GPT \u2013 often produce quick responses, which \u2018is not how science works\u2019, he says. \u2018[Scientific thinking] is slower, it\u2019s much more rigorous, deliberate, structured, and it\u2019s a process that often happens for a much longer period of time.\u2019<\/p>\n<p>To combat this challenge, Google DeepMind developed <a title=\"Co-Scientist | Google DeepMind\" href=\"https:\/\/deepmind.google\/blog\/co-scientist-a-multi-agent-ai-partner-to-accelerate-research\/?utm_source=keyword&amp;utm_medium=referral&amp;utm_campaign=geminiforscience&amp;utm_content=scienceblog\" rel=\"nofollow noopener\" target=\"_blank\">Co-Scientist<\/a> \u2013 an AI assistant purposefully built to work alongside scientists. DeepMind released the programme in February last year <a title=\"Towards an AI co-scientist | arXiv\" href=\"https:\/\/arxiv.org\/abs\/2502.18864\" rel=\"nofollow noopener\" target=\"_blank\">in a preprint<\/a>. Non-profit Futurehouse unveiled its own version, Robin, <a title=\"Robin: A multi-agent system for automating scientific discovery | arXiv\" href=\"https:\/\/arxiv.org\/abs\/2505.13400\" rel=\"nofollow noopener\" target=\"_blank\">a few months later<\/a> and both companies have now published their findings in Nature.1,2<\/p>\n<p>Co-Scientist builds on Google\u2019s AI assistant Gemini and uses several AI programmes or \u2018agents\u2019 to respond to a research goal proposed by a user. The agents then generate initial ideas, search existing scientific literature to evaluate hypotheses and iterate this process several times to refine ideas. This is similar to how Google\u2019s AlphaGo decides which move to make next in the board game Go, says Natarajan. Scientists are then able to carry out the suggested experiments, using the results to get Co-Scientist to generate better hypotheses.<\/p>\n<p>What is AI?<\/p>\n<p data-prosemirror-content-type=\"node\" data-prosemirror-node-name=\"paragraph\" data-prosemirror-node-block=\"true\" data-pm-slice=\"1 1 []\">Artificial intelligence (AI) is an umbrella term often incorrectly used to encompass a variety of connected but simpler processes.<\/p>\n<p data-prosemirror-content-type=\"node\" data-prosemirror-node-name=\"paragraph\" data-prosemirror-node-block=\"true\">AI is the ability of machines and computer programmes to perform tasks that typically only humans could do, such as reasoning, responding to feedback and decision making.<\/p>\n<p data-prosemirror-content-type=\"node\" data-prosemirror-node-name=\"paragraph\" data-prosemirror-node-block=\"true\">Generative AI is a newer variant of AI that analyses and detects patterns in training datasets to generate original text, images and videos in response to requests from users. ChatGPT, Microsoft Copilot, Google Gemini and more recently X\u2019s Grok are all examples of chatbots that use generative AI.<\/p>\n<p data-prosemirror-content-type=\"node\" data-prosemirror-node-name=\"paragraph\" data-prosemirror-node-block=\"true\">Neural networks are an interconnected array of artificial neurons, akin to biological brains, that identify, analyse and learn from statistical patterns in data.<\/p>\n<p data-prosemirror-content-type=\"node\" data-prosemirror-node-name=\"paragraph\" data-prosemirror-node-block=\"true\">Machine learning is a subset of AI that allows machines to learn from datasets and make predictions based on new data, without programmers explicitly asking it to do so. Machine learning models improve their performance as they receive more data.<\/p>\n<p data-prosemirror-content-type=\"node\" data-prosemirror-node-name=\"paragraph\" data-prosemirror-node-block=\"true\">Deep learning is an enhanced type of machine learning that uses neural networks with many layers to analyse complex data from very large datasets. Applications of deep learning include speech recognition, image generation and translation.<\/p>\n<p data-prosemirror-content-type=\"node\" data-prosemirror-node-name=\"paragraph\" data-prosemirror-node-block=\"true\">Large language models or LLMs are a type of deep learning trained on large amounts of data to understand and generate language. LLMs learn patterns in text by predicting the next word in the sequence and these models are now able to write prose, analyse text from the internet and hold dialogues with users.<\/p>\n<p>Natarajan explains that the team has worked with hundreds of scientists across the globe to see how the scientific community might use the tool. \u2018Most of the scientists [we speak to] are quite surprised at where AI is, what it is coming up with [and] how fast it is responding,\u2019 he says.<\/p>\n<p>For example, the team worked with microbiologists at Imperial College London, UK, who were researching <a title=\"AI mirrors experimental science to uncover a mechanism of gene transfer crucial to bacterial evolution | Cell\" href=\"https:\/\/www.cell.com\/cell\/fulltext\/S0092-8674(25)00973-0\" rel=\"nofollow noopener\" target=\"_blank\">how certain genes are transferred between bacteria that allow them to develop antimicrobial resistance<\/a>. \u2018When we ran the [Co-Scientist] system for a couple of days, it essentially recapitulated their entire research journey and made the exact same predictions,\u2019 says Natarajan.<\/p>\n<p><a title=\"Tiago Dias da Costa | Imperial\" href=\"https:\/\/profiles.imperial.ac.uk\/t.costa\" rel=\"nofollow noopener\" target=\"_blank\">Tiago Dias da Costa<\/a>, who led the research at Imperial, says that Co-Scientist had \u2018independently generated a hypotheses that closely matched the mechanism we had uncovered through years of experimental work, but \u2018it did not replace the experimental discovery process\u2019.<\/p>\n<p>\u2018The value of Co-Scientist was not that it simply \u201cgave us the answer\u201d but that it demonstrated how AI can help generate, prioritise and refine biologically meaningful hypotheses\u2019. He stresses that \u2018AI-generated hypotheses only become discoveries when they are tested rigorously in the lab\u2019.<\/p>\n<p>Co-Scientist will become the hypothesis-generating tool within Google\u2019s broader <a title=\"Gemini for Science | Google\" href=\"https:\/\/ai.google\/gemini-for-science\/\" rel=\"nofollow noopener\" target=\"_blank\">Gemini for Science<\/a> programme, which will be available to researchers in \u2018the next few weeks and months\u2019, he adds. However, Natarajan thinks that \u2018there\u2019s still a few leaps that are needed before we have a system that can do what some of the great scientists of the past [have done] \u2013 like coming up with a true original breakthrough [or] paradigm shifting theory\u2019.<\/p>\n<p>Repurposing existing drugs to find new ways to treat disease<\/p>\n<p>Futurehouse\u2019s Robin works in a similar way to Co-Scientist. \u2018A person can just input the name of a disease \u2013 any disease and no other context \u2013 and Robin will extensively search the literature and come up with experiments that model different components,\u2019 says <a title=\"Michaela Hinks | LinkedIn\" href=\"https:\/\/www.linkedin.com\/in\/michaela-hinks-729103129\/\" rel=\"nofollow noopener\" target=\"_blank\">Michaela Hinks<\/a>, who helped develop Futurehouse\u2019s programme.<\/p>\n<p>She adds that Robin will suggest existing drugs that scientists can then test in the lab. Inputting the experimental data into Robin allows the system to generate a refined set of hypotheses. What makes Robin stand out, says Hinks, is that the system only uses three agents. She explains that other similar tools were \u2018just adding complexity for the sake of being fancy\u2019. \u2018I think systems should be as simple as possible to do the job.\u2019<\/p>\n<p>The team tested Robin by attempting to find treatments for dry age-related macular degeneration (dAMD) \u2013 a sight loss condition that affects millions of people in the US, which currently has no cure.<\/p>\n<p>Hinks explains that Robin generated a novel way to treat the disease by enhancing the destruction of certain eye cells. Robin then proposed an experiment to measure the destruction of these cells and molecules that might treat the condition, with biological tests revealing two possibilities. This included Ripasudil \u2013 a clinically approved drug to treat glaucoma \u2013 and a circadian clock modulator, which was an unexpected result.<\/p>\n<p>\u2018I think that this was probably a good testbed to use for this system to see if it could recapitulate the fact that human scientists have actually thought about doing this,\u2019 says <a title=\"Derek Lowe | Chemistry World\" href=\"https:\/\/www.chemistryworld.com\/derek-lowe\/1294.bio\" rel=\"nofollow noopener\" target=\"_blank\">Derek Lowe<\/a>, a US-based drug discovery chemist and Chemistry World columnist. He adds that when he searched for scientific papers with Ripasudil and dAMD, he found lots of results.<\/p>\n<p>Hinks adds that Robin generates hypotheses by accessing the open-source literature \u2018to make sure what the model proposes is grounded in what we already know\u2019.<\/p>\n<p>\u2018[Being able to generate hypotheses] is only going to be useful in cases where you have a lot of very good, reliable data to feed into these things,\u2019 says Lowe, \u2018because otherwise it\u2019s going to spit out all kinds of stuff that\u2019s not going to pan out. But join the club, we humans do that too,\u2019 he adds.<\/p>\n<p>\u2018Any machine learning system is absolutely dependent on the highest quality data that you can give it \u2013 both positive and negative results \u2013 and experimental observations as well,\u2019 says Lowe. He adds that the scientific literature is currently \u2018a bit of a mess\u2019, adding that the influx of research papers \u2013 increasingly ones that include AI-generated results or text \u2013 will not help the situation. Lowe thinks that these tools will not be able to answer fundamental scientific questions. \u2018You\u2019re not going to get all the answers by doing this because that would presuppose that all the answers are there in the literature.\u2019<\/p>\n<p>\u2018I don\u2019t think that your standard academic or industrial drug discovery scientists are going to start using [these tools] immediately,\u2019 says Lowe. \u2018These are the equivalent, to me, of a laboratory instrument that still has soldered wires and duct tape and things hanging out of it.\u2019 He thinks that things have a long way to go before they are ready for a regular end user to benefit from such tools.<\/p>\n","protected":false},"excerpt":{"rendered":"Artificial intelligence (AI) research labs Google DeepMind and Futurehouse have now released AI research assistants that can generate&hellip;\n","protected":false},"author":2,"featured_media":462223,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[7],"tags":[111,139,69,147],"class_list":["post-462222","post","type-post","status-publish","format-standard","has-post-thumbnail","category-science","tag-new-zealand","tag-newzealand","tag-nz","tag-science"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/posts\/462222","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/comments?post=462222"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/posts\/462222\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/media\/462223"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/media?parent=462222"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/categories?post=462222"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/tags?post=462222"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}