"The lab was run by scientists in white coats, automated equipment, and invisible software," wrote Steve Lorre of the New York Times, following a visit to the Science Night Society.
At "Science Night," they're trying to reinvent the scientific process itself. And at least for now, they seem to be succeeding. With the help of artificial intelligence, of course.
If you enter the six-story building of Science Night, a company that recently received a $96 million investment, you will discover a magical laboratory. Trays containing XNUMX lions, each with its own separate DNA sample, are transported rapidly on magnetic tracks and move from station to station in the laboratory. At each station they receive different treatment, sometimes by humans, sometimes by machines, before being passed on to the next station.
And all under the guidance of artificial intelligence, which rules the laboratory with a high hand. Alongside the human professors, at least for now.
“The lab was run by scientists in white coats, automated equipment and invisible software,” wrote Steve Lorre of The New York Times, following a visit to the building. “Every measurement, every experiment, every gradual success and failure was digitally recorded and fed into Lila’s artificial intelligence. So she was always learning, getting smarter and doing more on her own.”
The ultimate goal? As one of the founders said in a quote from that article –
"Our goal is to really give AI access to running the scientific method: coming up with new ideas and actually going into the lab and testing those ideas."
This is the ultimate vision. Artificial intelligence that not only augments – or replaces – the minds of scientists, but also the experiments they do in the lab.
But let's start at the beginning.
The Great Competition of the 21st Century
In 1942, the United States launched the famous Manhattan Project, which brought together the best minds for the most terrible goal: creating a bomb that could destroy entire cities. The efforts of those scientists led to the development of the first atomic bomb, which was dropped on Hiroshima and Nagasaki and cost the lives of hundreds of thousands of Japanese. It is no exaggeration to say that if the Germans had succeeded in developing the atomic bomb before the Americans, World War II could have ended with Hitler standing over the radioactive ruins of Washington.
This is the power of science. To heal, to build – and also to destroy. The country that enjoys the most advanced science and technology can start and end wars with the push of a button.
In late 2024, the US Congressional Committee on Relations with China issued an extraordinary recommendation: to launch a new Manhattan Project. That modern Manhattan Project would not focus on developing a new atomic bomb, but its power to harm and benefit may be even greater.
At the end of the new Manhattan Project, will the United States succeed in developing artificial intelligence that is as smart as the smartest scientists today – and even smarter.
The United States government has not yet officially opened that Manhattan Project, perhaps due to Warnings from people like the former Google CEO, according to which a declaration of the start of the AI race could cause many other countries to abandon all the destroyers and obstacles in their path to developing superintelligence. And yet, one of Trump's first steps in power – just one day after his inauguration – was the launch of the "Stargate AI Project". The project's goal is not to create superintelligence explicitly, but 'only' to recruit andInvest five hundred billion dollars For the next four years in artificial intelligence.
But why? Why invest half a trillion dollars in artificial intelligence?
Simple: Trump, like Xi Jinping and all world leaders who pay attention to the latest trends, understand that real power comes from science and technology, which are the product of the minds of scientists. The country that succeeds in upgrading its scientists – or replacing them – with artificial intelligence will gain immense power in every field.
Power to invent new drugs that will slow aging and reduce mortality from deadly diseases.
Power to purify water and grow crops at a greater rate than ever before, and feed all of the country's citizens – and sell the rest to other countries.
Power to develop technologies that will harvest energy from the environment or from nuclear fusion, and provide all citizens with electricity at zero cost.
And also the power to invent and perfect more effective weapons than ever before.
This is the power that scientists give to countries. This is the power that artificial intelligence will give them, when it reaches the level of the human scientist and can replace him in his work.
Well, what does it take to replace a scientist?
How does a scientist do it?
August 2024 Sakana AI has announced the first “artificial scientist.” Sakana has created an artificial intelligence that can go through all the steps of an average scientist. It reads scientific studies, identifies gaps in current scientific knowledge, develops hypotheses that explain the gaps, suggests experiments to refute the hypotheses – and then performs the experiments, collects the results, and summarizes conclusions. At the end of the process, it even writes the scientific article that describes the entire study.
There's just one problem: Sakena did this entire process in a virtual world, and only for computer science research. Why? Because she didn't have any robots.
One critical step in any scientific study is coming up with hypotheses—that is, new ideas that explain reality. But hypotheses are like opinions: everyone has one. So the second critical part of the study is testing the hypotheses in the lab. Sakena was able to do this for computer science research because her artificial intelligence suggested experiments that could be run on the computer itself. There was no need for an actual lab.
As a side note, we would like to add that last week, Sakena Managed to publish the first article Its artificial intelligence product, at one of the leading conferences in the field of artificial intelligence. The paper underwent careful peer review, and received a score that was higher than that of 45 percent of the other papers submitted to the conference. In other words, in certain fields it is possible to produce new scientific knowledge even without conducting actual laboratory experiments. Still, those who want to advance the life sciences, materials engineering, chemistry, physics, and almost any other field – need to find a way to connect the brain of artificial intelligence to the hands and feet of scientists in the laboratory.
And this is where "Science Night" came into the picture.
Brain and body – in one laboratory
"Science Night" was founded following several projects in which it demonstrated how its artificial intelligence is able to carry out the research process alongside humans, and in combination with autonomous laboratories. As part of the projects, scientists explained to the artificial intelligence what they wanted it to accomplish – and it suggested ideas and experiments to test them. The experiments were not only carried out by humans. As Which is described on the "Science Night" website, the experiments were transferred to –
"An autonomous science platform designed to expand and optimize experiments in any scientific field under human guidance, by combining creative artificial intelligence with autonomous science units..."
According to the company, it has managed to achieve impressive performance in a variety of areas in the past three years. Among its successes, it lists the creation of "optimal structures for genetic medicine that achieve higher performance than currently available treatments on the market," or "the creation of a unique group of non-platinum metals that will be used as catalysts in the production of green hydrogen, at a much lower cost than catalysts available on the market," or "the design of a new material for carbon adsorption on an industrial scale, with better capacity, thermal stability, and kinetic adsorption ability than leading products."
All of these claims, as of now, are not supported by scientific articles that have reached the scientific press and have been examined by independent experts in the field. We should, in fact, rely only on the promises of "Science Night" itself. And yet, the company has on its staff leading scientists such as George Church, who is considered the "modern-day Edison" in the field of genetic engineering. It has also managed to convince investors to pour two hundred million dollars into its bank account - and such an investment suggests that there is something amiss.
If I had to guess, it would be that the investors were probably exposed to the vision that "Science Night" presents on its website. And it's a truly grand and ambitious vision - the kind that could almost be called the "Manhattan Project of artificial intelligence."
“We believe the next frontier in scientific discovery will come from enabling artificial intelligence to design and run experiments on an unprecedented scale. … With every advance in the scientific method, AI systems will be able to coordinate thousands of experiments simultaneously, systematically building our understanding and accelerating scientific progress,” wrote Dr. Molly Gibson, co-founder and president of Science Night. She added that “this approach has the potential to advance research in many scientific fields, from health and drug development to sustainable agriculture and resources.”
Will "Science Night" be successful? It's hard to know in advance. But one thing is certain: artificial intelligence is already beginning to assist scientists in their work, and it's easy to assume that with the right adjustments and the right combinations of a computer brain and sophisticated machines, it will be able to design and run "thousands of experiments simultaneously" in certain fields.
And what will be the impact on the world?
From Marx to the present day
The thinker Karl Marx (yes, the one who co-invented communism) once wrote that popular culture always serves the economy. Why do we value innovation and creativity so much these days? Because one innovative and ambitious person can reinvent an entire technological field, making a huge contribution to industry and the economy. Why do we value the wisdom and resourcefulness of scientists? For the same reason.
So, what value will scientists have – for all their human wisdom – in a world where artificial intelligence can conduct experiments on a scale that leaves them far behind?
Would you say that even in such a world, scientists are still needed to monitor artificial intelligence? True, but what happens when artificial intelligence can monitor itself better than any scientist? When it can understand on its own which experiments are more and less suitable for the hypotheses it raises? When it can even improve itself thanks to the scientific and technological insights it produces every year?
This vision may sound like science fiction, but it could be a reality within the next decade. It requires the development of “artificial general intelligence”—that is, one that reaches the level of human scientists who are experts in any field. It turns out that the prediction markets, where people bet real money on predictions of the future, believe that the chance of such progress is 62 percent for 2030, and 75 percent by 2035. Given these odds, we need to start discussing these questions very seriously.
Risk and opportunity
In the short story "Autopeck," science fiction writer Philip K. Dick described the factory of the future. Its name, of course, is Autopeck, a portmanteau of the words "automaton" and "factory." This factory never stops working. Robots do all the work, and they can repair themselves if necessary. If resources run out, the robots can send machines to find new resources and bring them to the factory to continue producing.
The idea is excellent in itself, but Dick – as is the custom of science fiction writers – takes it to the extreme, and then a little beyond. In the full story, humans lose control of Autopak after a great war. The automated factory continues to produce milk and all kinds of food for them, but at the same time, it consumes the other resources they need to survive. The humans try to convince the factory to stop its operations, but it is not convinced. In the end, they realize that the only way to disable it is to pit it against other Autopaks in a war for the resources it needs. The Great Autopak War begins, and at the end, the humans discover that the most innovative Autopak has found a way to avoid destruction: it launches small capsules containing all the machines necessary to establish new Autopaks all over the world.
Dick does describe an extreme situation in which machines get out of control, but as mentioned – science fiction has begun to come true in recent years with increasing speed. It is not difficult to imagine a system like that of "Science Night" that becomes autonomous and provides mankind with inventions of all kinds, with human supervision that only comes at the highest level. That is, politicians would define for artificial intelligence what is necessary for the country – and it would conduct the appropriate research to achieve what is desired.
Medicines for everyone? Plenty. Fertilizers and genetically modified plants? For everyone. Advanced explosives, biological weapons, and nuclear bombs that can fit in a shirt pocket? Why not.
The opportunity is enormous, and so is the risk.
When the original Manhattan Project ended, humanity discovered the vast and terrifying power of the atom. Alongside it, it also discovered the enormous potential for producing clean and cheap energy. Similarly, the new scientific project in the field of artificial intelligence will certainly reveal new dangers, but also great opportunities. The difference this time is that we are entering an era in which technology itself learns, develops and leads itself forward. This is a new world, and the implications are not yet completely clear – but it is already clear that the future will belong to the country that succeeds in harnessing the power of scientific artificial intelligence to its advantage.
And one can hope for the benefit of all humanity as well.
