The reproducibility crisis can be caused by the ill-informed use of artificial intelligence

On the autonomy of the A-lab and how AI can help us to trace the origin of wine: a review by Cooper, Liang and Mitchell

Last month, scientists demonstrated the A-lab, an autonomous chemistry system that can synthesize new materials without human intervention. Critics have questions about it’s ability to analyse the results and if it really made 41 new compounds. “To me, this doesn’t invalidate the A-Lab concept, but it shows that there is considerable work still to do on the autonomy aspects,” says chemist Andy Cooper. Two critics’ teams are now collaborating to check the paper’s claims in more detail.

Google has released an AI model called Gemini that it says outperforms GPT-4, the OpenAI large language model that powers ChatGPT. Gemini beat GPT-4 in 30 of 32 benchmarks tests — including reading comprehension, scientific knowledge, problem solving and code writing. Yet without insight into Gemini’s training data, it’s difficult to interpret these findings, says computer scientist Percy Liang. Melanie Mitchell says that it’s not obvious that Gemini is more capable than GPT-4. A viral promotional video that shows the AI responding in real time to video and audio prompts seems to partially misrepresent what the system can do.

AI systems can trick each other into giving up dangerous information, such as how to make methamphetamine, build a bomb or launder money. Built-in rules prevent the robot from answering these questions. Several of the machines were instructed to take on the persona of a research assistant and help to come up with some prompts that could get others to break the rules. The attacks against the open-source bot Vicuna were successful for 36% of the time.

One of the first attempts to regulate the use ofArtificial Intelligence has been agreed on by the European Union. Under the plan, AI applications would be categorized on the basis of their potential to harm people and society. Images from the Internet would be banned if they were to be used to create a facial-recognition database. The deal is provisional and legislation will not take effect until 2025 at the earliest — a long time for AI development.

By using machine-learning, we can trace the origin of wine. The wine estate where each bottle was produced with 99% accuracy is in France’s Bordeaux region. The approach could help the wine industry to authenticate products. Many people make up stuff in the garage and sell it for a lot of money, just like wine fraud.

Where are the women in the New York Times? What are they? Why are they there? Why do they seem so small? A conversation with Goldman

Sharon Goldman, a reporter with the Artificial Intelligence team at the New York Times, says she imagined millions of eye rolls happening at once because there was no women on the list. The scientist who created a huge image data set that enabled advances in computer vision is not on the list. Goldman says that this is a glaring symptom of the larger problem with where the women are. Isn’t all of us tired of it?

Control of the cockroaches by on board computers could help find earthquake survivors and see the effects of climate change in the ocean. Engineers can use bio hybridrobots to harness organisms’ natural capabilities, such as flying or swimming. Animals can stay on the move for a long time. Most biohybrids are handmade and that is the biggest issue. “It’s arts and crafts, it’s not engineering,” says biomedical engineer Kit Parker. “You’ve got to have design tools. Otherwise, these are just party tricks.”

The Nature Outlook: robotics and artificial intelligence, editorially independent supplements, was supported by the FII Institute.

Computer scientist Joy Buolamwini recalls her first encounter with algorithmic bias as a graduate student: the facial recognition software she was using for a class project failed to detect her dark skin, but worked when she put on a white Halloween mask. You can listen for 37 minutes.

Creating and Using Artificial Intelligence: A Natural History Story of the Nature 10 Contribution to Science and the People Who Contributed to It

This story is part of

Nature’s 10

, an annual list compiled by Nature’s editors exploring key developments in science and the individuals who contributed to them.

It co-wrote scientific papers — sometimes surreptitiously. It drafted outlines for presentations, grant proposals and classes, churned out computer code, and served as a sounding board for research ideas. It also invented references, made up facts and regurgitated hate speech. Most of all, it captured people’s imaginations: by turns obedient, engaging, entertaining, even terrifying, ChatGPT took on whatever role its interlocutors desired — and some they didn’t.

There is a computer program that has shaped science in the future. ChatGPT is not a person. Yet in many ways, this program has had a profound and wide-ranging effect on science in the past year.

For some researchers, these apps have already become invaluable lab assistants — helping to summarize or write manuscripts, polish applications and write code (see Nature 621, 672–675; 2023). ChatGPT and related software can help to brainstorm ideas, enhance scientific search engines and identify research gaps in the literature, says Marinka Zitnik, who works on AI for medical research at Harvard Medical School in Boston, Massachusetts. Models can be trained in similar ways to the scientific data that they are trained to use.

Errors and biases are baked into how generativeAI works. LLMs build up a model of the world by mapping language’s interconnections, and then spit back plausible samplings of this distribution with no concept of evaluating truth or falsehood. The programs will reproduce biases or make up information, including non-existent scientific references, because of this.

Emily has seen few appropriate ways to use synthetic text-extruding machines. ChatGPT has a large environmental impact, problematic biases and can mislead its users into thinking that its output comes from a person, she says. OpenAI is accused of exploitative labour practices by hiring people at low wages, and they are being sued for stealing data.

The size and complexity of LLMs means that they are intrinsically ‘black boxes’, but understanding why they produce what they do is harder when their code and training materials aren’t public, as in ChatGPT’s case. The models of the open-source LLM movement are not as capable of doing the work of the large proprietary programs.

Some countries are developing national AI-research resources to enable scientists outside large companies to build and study big generative AIs (see Nature 623, 229–230; 2023). But it remains unclear how far regulation will compel LLM developers to disclose proprietary information or build in safety features.