Singularity HUBModify

2026-01-30 06:46:16

DNA stores the body’s operating playbook. Some genes encode proteins. Other sections change a cell’s behavior by regulating which genes are turned on or off. For yet others, the dark matter of the genome, the purpose remains mysterious—if they have any at all.

Normally, these genetic instructions conduct the symphony of proteins and molecules that keep cells humming along. But even a tiny typo can throw molecular programs into chaos. Scientists have painstakingly connected many DNA mutations—some in genes, others in regulatory regions—to a range of humanity’s most devastating diseases. But a full understanding of the genome remains out of reach, largely because of its overwhelming complexity.

AI could help. In a paper published this week in Nature, Google DeepMind formally unveiled AlphaGenome, a tool that predicts how mutations shape gene expression. The model takes in up to one million DNA letters—an unprecedented length—and simultaneously analyzes 11 types of genomic mutations that could torpedo the way genes are supposed to function.

Built on a previous iteration called Enformer, AlphaGenome stands out for its ability to predict the purpose of DNA letters in non-coding regions of the genome, which largely remain mysterious.

Computational gene expression prediction tools already exist, but they’re usually tailored to one type of genetic change and its consequences. AlphaGenome is a jack-of-all-trades that tracks multiple gene expression mechanisms, allowing researchers to rapidly capture a comprehensive picture of a given mutation and potentially speed up therapeutic development.

Since its initial launch last June, roughly 3,000 scientists from 160 countries have experimented with the AI to study a range of diseases including cancer, infections, and neurodegenerative disorders, said DeepMind’s Pushmeet Kohli in a press briefing.

AlphaGenome is now available for non-commercial use through a free online portal, but the DeepMind team plans to release the model to scientists so they can customize it for their research.

“We see AlphaGenome as a tool for understanding what the functional elements in the genome do, which we hope will accelerate our fundamental understanding of the code of life,” said study author Natasha Latysheva in the news conference.

98 Percent Invisible

Our genetic blueprint seems simple. DNA consists of four basic molecules represented by the letters A, T, C, and G. These letters are grouped in threes called codons. Most codons call for the production of an amino acid, a type of molecule the body strings together into proteins. Mutations thwart the cell from making healthy proteins and potentially cause diseases.

The actual genetic playbook is far more complex.

When scientists pieced together the first draft of the human genome in the early 2000s, they were surprised by how little of it directed protein manufacturing. Just two percent of our DNA encoded proteins. The other 98 percent didn’t seem to do much, earning the nickname “junk DNA.”

Over time, however, scientists have realized those non-coding letters have a say about when and in which cells a gene is turned on. These regions were originally thought to be physically close to the gene they regulated. But DNA snippets thousands of letters away can also control gene expression, making it tough to hunt them down and figure out what they do.

It gets messier.

Cells translate genes into messenger molecules that shuttle DNA instructions to the cell’s protein factories. In this process, called splicing, some DNA sequences are skipped. This lets a single gene create multiple proteins with different purposes. Think of it as multiple cuts of the same movie: The edits result in different but still-coherent storylines. Many rare genetic diseases are caused by splicing errors, but it’s been hard to predict where a gene is spliced.

Then there’s the accessibility problem. DNA strands are tightly wrapped around a protein spool. This makes it physically impossible for the proteins involved in gene expression to latch on. Some molecules dock onto tiny bits of DNA and tug them away from the spool to provide access, but the sites are tough to hunt down.

The DeepMind team thought AI would be well-suited to take a crack at these problems.

“The genome is like the recipe of life,” said Kohli in a press briefing. “And really understanding ‘What is the effect of changing any part of the recipe?’ is what AlphaGenome sort of looks at.”

Making Sense of Nonsense

Previous work linking genes to function inspired AlphaGenome. It works in three steps. The first detects short patterns of DNA letters. Next the algorithm communicates this information across the entire analyzed DNA section. In the final step, AlphaGenome maps detected patterns into predictions like, for example, how a mutation affects splicing.

The team trained AlphaGenome on a variety of publicly available genetic libraries amassed by biologists over the past decade. Each captures overlapping aspects of gene expression, including differences between cell types and species. AlphaGenome can analyze sequences that are as long as a million DNA letters from humans or mice. It can then predict a range of molecular outcomes at the resolution of single letter changes.

“Long sequence context is important for covering regions regulating genes from far away,” wrote the team in a blog post. The algorithm’s high resolution captures “fine-grained biological details.” Older methods often sacrifice one for the other; AlphaGenome optimizes both.

The AI is also extremely versatile. It can make sense of 11 different gene regulation processes at once. When pitted against state-of-the-art programs, each focused on just one of these processes, AlphaGenome was as good or better across the board. It readily detected areas engaged in splicing and scored how much DNA letter changes would likely affect gene expression.

In one test, the AI tracked down DNA mutations roughly 8,000 letters away from a gene involved in blood cancer. Normally, the gene helps immune cells mature so they can fight off infections. Then it turns off. But mutations can keep it switched on, causing immune cells to replicate out of control and turn cancerous. That the AI could predict the impact of these far-off DNA influences showcases its genome-deciphering potential.

There are limitations, however. The algorithm struggles to capture the roles of regulatory regions over 100,000 DNA letters away. And while it can predict molecular outcomes of mutations—for example, what proteins are made—it can’t gauge how they cause complex diseases, which involve environmental and other factors. It’s also not set up to predict the impact of DNA mutations for any particular individual.

Still, AlphaGenome is a baseline model that scientists can fine-tune for their area of research, provided there’s enough well-organized data to further train the AI.

“This work is an exciting step forward in illuminating the ‘dark genome.’ We still have a long way to go in understanding the lengthy sequences of our DNA that don’t directly encode the protein

machinery whose constant whirring keeps us healthy,” said Rivka Isaacson at King’s College London, who was not involved in the work. “AlphaGenome gives scientists whole new and vast datasets to sift and scavenge for clues.”

The post Google DeepMind’s AlphaGo Decodes the Genome a Million ‘Letters’ at a Time appeared first on SingularityHub.

2026-01-28 09:06:17

Creativity is a trait that AI critics say is likely to remain the preserve of humans for the foreseeable future. But a large-scale study finds that leading generative language models can now exceed the average human performance on linguistic creativity tests.

The question of whether machines can be creative has gained new salience in recent years thanks to the rise of AI tools that can generate text and images with both fluency and style. While many experts say true creativity is impossible without lived experience of the world, the increasingly sophisticated outputs of these models challenge that idea.

In an effort to take a more objective look at the issue, researchers at the Université de Montréal, including AI pioneer Yoshua Bengio, conducted what they say is the largest ever comparative evaluation of machine and human creativity to date. The team compared outputs from leading AI models against responses from 100,000 human participants using a standardized psychological test for creativity and found that the best models now outperform the average human, though they still trail top performers by a significant margin.

“This result may be surprising—even unsettling—but our study also highlights an equally important observation: even the best AI systems still fall short of the levels reached by the most creative humans,” Karim Jerbi, who led the study, said in a press release.

The test at the heart of the study, published in Scientific Reports, is known as the Divergent Association Task and involves participants generating 10 words with meanings as distinct from one another as possible. The higher the average semantic distance between the words, the higher the score.

Performance on this test in humans correlates with other well-established creativity tests that focus on idea generation, writing, and creative problem solving. But crucially, it is also quick to complete, which allowed the researchers to test a much larger cohort of humans over the internet.

What they found was striking. OpenAI’s GPT-4, Google’s Gemini Pro 1.5 and Meta’s Llama 3 and Llama 4, all outperformed the average human. However, when they measured the average performance of the top 50 percent of human participants, it exceeded all tested models. The gap widened further when they took the average of the top 25 percent and top 10 percent of humans.

The researchers wanted to see if these scores would translate to more complex creative tasks, so they also got the models to generate haikus, movie plot synopses, and flash fiction. They analyzed the outputs using a measure called Divergent Semantic Integration, which estimates the diversity of ideas integrated into a narrative. While the models did relatively well, the team found that human-written samples were still significantly more creative than AI-written ones.

However, the team also discovered they could boost the AI’s creativity with some simple tweaks. The first involved adjusting a model setting called temperature, which controls the randomness of the model’s output. When this was turned all the way up on GPT-4, the model exceeded the creativity scores of 72 percent of human participants.

The researchers also found that carefully tuning the prompt given to the model helped too. When explicitly instructed to use “a strategy that relies on varying etymology,” both GPT-3.5 and GPT-4 did better than when given the original, less-specific task prompt.

For creative professionals, Jerbi says the persistent gap between top human performers and even the most advanced models should provide some reassurance. But he also thinks the results suggest  people should take these models seriously as potential creative collaborators.

“Generative AI has above all become an extremely powerful tool in the service of human creativity,” he says. “It will not replace creators, but profoundly transform how they imagine, explore, and create—for those who choose to use it.”

Either way, the study adds to a growing body of research that is raising uncomfortable questions about what it means to be creative and whether it is a uniquely human trait. Given the strength of feeling around the issue, the study is unlikely to settle the matter, but the findings do mark one of the more concrete attempts to measure the question objectively.

The post AI Now Beats the Average Human in Tests of Creativity appeared first on SingularityHub.

2026-01-25 23:00:00

Stuck in front of our screens all day, we often ignore our senses beyond sound and vision. And yet they are always at work. When we’re more alert we feel the rough and smooth surfaces of objects, the stiffness in our shoulders, the softness of bread.

In the morning, we may feel the tingle of toothpaste, hear and feel the running water in the shower, smell the shampoo, and later the aroma of freshly brewed coffee.

Aristotle told us there were five senses. But he also told us the world was made up of five elements, and we no longer believe that. And modern research is showing we may actually have dozens of senses.

Almost all of our experience is multisensory. We don’t see, hear, smell, and touch in separate parcels. They occur simultaneously in a unified experience of the world around us and of ourselves.

What we feel affects what we see, and what we see affects what we hear. Different odors in shampoo can affect how you perceive the texture of hair. The fragrance of rose makes hair seem silkier, for instance.

Odors in low-fat yogurts can make them feel richer and thicker on the palate without adding more emulsifiers. Perception of odors in the mouth, rising to the nasal passage, are modified by the viscosity of the liquids we consume.

My long-term collaborator, professor Charles Spence from the Crossmodal Laboratory in Oxford, told me his neuroscience colleagues believe there are anywhere between 22 and 33 senses.

These include proprioception, which enables us to know where our limbs are without looking at them. Our sense of balance draws on the vestibular system of ear canals as well as sight and proprioception.

Another example is interoception, by which we sense changes in our own bodies such as a slight increase in our heart rate and hunger. We also have a sense of agency when moving our limbs: a feeling that can go missing in stroke patients who sometimes even believe someone else is moving their arm.

There is the sense of ownership. Stroke patients sometimes feel their, for instance, arm is not their own even though they may still feel sensations in it.

Some of the traditional senses are combinations of several senses. Touch, for instance involves pain, temperature, itch, and tactile sensations. When we taste something, we are actually experiencing a combination of three senses: touch, smell, and taste—or gustation—which combine to produce the flavors we perceive in food and drinks.

Gustation, covers sensations produced by receptors on the tongue that enable us to detect salt, sweet, sour, bitter, and umami (savory). What about mint, mango, melon, strawberry, raspberry?

We don’t have raspberry receptors on the tongue, nor is raspberry flavor some combination of sweet, sour, and bitter. There is no taste arithmetic for fruit flavors.

We perceive them through the combined workings of the tongue and the nose. It is smell that contributes the lion’s share to what we call tasting.

This is not inhaling odors from the environment, though. Odor compounds are released as we chew or sip, traveling from the mouth to the nose though the nasal pharynx at the back of throat.

Touch plays its part too, binding tastes and smells together and fixing our preferences for runny or firm eggs and the velvety, luxurious gooeyness of chocolate.

Sight is influenced by our vestibular system. When you are on board an aircraft on the ground, look down the cabin. Look again when you are in the climb.

It will “look” to you as though the front of the cabin is higher than you are, although optically, everything is in the same relation to you as it was on the ground. What you “see” is the combined effect of sight and your ear canals telling you that you are titling backwards.

The senses offer a rich seam of research and philosophers, neuroscientists and psychologists work together at the Center for the Study of the Senses at the University of London’s School of Advanced Study.

In 2013, the center launched its Rethinking the Senses project, directed by my colleague, the late Professor Sir Colin Blakemore. We discovered how modifying the sound of your own footsteps can make your body feel lighter or heavier.

We learned how audioguides in Tate Britain art museum that address the listener as if the model in a portrait was speaking enable visitors to remember more visual details of the painting. We discovered how aircraft noise interferes with our perception of taste and why you should always drink tomato juice on a plane.

While our perception of salt, sweet, and sour is reduced in the presence of white noise, umami is not, and tomatoes and tomato juice are rich in umami. This means the aircraft’s noise will taste enhance the savory flavor.

At our latest interactive exhibition, Senses Unwrapped at Coal Drops Yard in London’s King’s Cross, people can discover for themselves how their senses work and why they don’t work as we think they do.

For example, the size-weight illusion is illustrated by a set of small, medium, and large curling stones. People can lift each one and decide which is heaviest. The smallest one feels heaviest, but people can then place them on balancing scales and discover that they are all the same weight.

But there are always plenty of things around you to show how intricate your senses are, if you only pause for a moment to take it all in. So next time you walk outside or savor a meal, take a moment to appreciate how your senses are working together to help you feel all the sensations involved.

This article is republished from The Conversation under a Creative Commons license. Read the original article.

The post Humans Could Have as Many as 33 Senses appeared first on SingularityHub.

2026-01-24 23:00:00

The post This Week’s Awesome Tech Stories From Around the Web (Through January 24) appeared first on SingularityHub.

2026-01-24 09:14:03

In the 1980s, UCLA cellular biologist Leonard Rome noticed odd, barrel-shaped structures present in almost all cells. The hollow particles were filled with RNA and a handful of proteins. Naming them vaults, Rome has tried to understand their purpose ever since.

Though vaults remain enigmatic, their unique structure recently inspired a separate team. Led by Fei Chen at the Broad Institute of MIT and Harvard, the scientists engineered vaults to collect and store messenger RNA (mRNA) molecules for up to a week. The mRNA vaults they created act like ledgers that detail which genes are turned on or off over time.

In several tests, opening the vaults and reading the mRNA stored within shed light on gene activity that helps cancer cells evade treatment. The method, called TimeVault, also tracked the intricate symphony of gene expression that pushes stem cells to mature into different cell types.

The work is “superpowerful” and “very innovative,” Jiahui Wu at the University of Massachusetts, who was not involved in the study, told Science.

Jay Shendure, an expert in cellular recorders at the University of Washington, agrees. It took “some creativity and some guts” to transform vaults into time capsules, he told Nature.

A Cell’s Life

Each cell is a metropolis humming with activity. Proteins zoom across its interior to coordinate behaviors. Structures called organelles churn out new proteins or recycle old ones to keep cells healthy. Scores of signaling molecules relay information from the environment to the nucleus, where our DNA resides. All this information causes the cell to turn certain genes on or off, allowing it to adapt to a changing biological world.

Scientists have long tried to spy on these intricate cellular processes. Using a common tool, they can tag molecules with glow-in-the-dark protein markers and track them under the microscope. This provides real-time data but only for a handful of proteins over a relatively short time.

Another approach takes snapshots of which genes are active in single cells or groups of cells, usually at the beginning and end of an experiment. Here, scientists extract mRNA, a molecule that carries gene expression information, to paint an overall picture of a cell’s current state. Comparing genetic activity between one point of time and another provides insight into the cell’s history. But unlike a video, these snapshots can’t capture nuanced changes over time.

More recently, a slew of cell recorders based on the gene editor CRISPR have galvanized the field. These tools encode information about cellular events into DNA, essentially forming a “video” of events inside cells that can be retrieved later by sequencing the DNA. Genomic recordings are relatively stable and have been used to map cell lineages—a bit like reconstructing a family tree—and record specific cell signals, such as those responding to viral infection, inflammation, nutrients, or other stimuli. But because they directly write into DNA, the process takes time and could trigger off-target effects.

Enter the Vault

Instead of tinkering with the genetic blueprint, mRNA may be a safer choice. These molecules carry protein-making instructions from DNA and have a relatively short lifespan. In other words, they reflect all the active genes in a cell at any moment, making them perfect candidates for a time capsule. But without protection, they’re rapidly destroyed—often within hours.

The team first tried to stabilize mRNA molecules by tethering them to a bacterial protein. It didn’t work. But after serendipitously stumbling across a YouTube channel by the Vault Guy, also known as Leonard Rome, they had an out-of-the-box idea. Cellular vaults are known to encapsulate some of life’s molecules. Could they also keep mRNA safe?

Vaults are made of 78 copies of a long protein. These proteins are woven into a barrel-shaped shell with a mostly hollow interior. To make their vault-based time capsule, the team first made a protective protein cap for the mRNA. This stabilized the molecules. The cap also links up with a slightly tweaked vault protein, engineered to tether captured mRNA molecules into a vault.

The team built in a switch too. TimeVault starts recording when cells are dosed with a chemical and stops as soon as the chemical washes out. Viewing the recording of gene activity is simple. The team retrieves the vaults and sequences all of the mRNA inside. TimeVault reliably stores the molecules for at least a week in multiple types of cells in petri dishes.

In a test, the technology faithfully captured mRNA in cells exposed to heat or low oxygen. Both are common ways to stress cells and force them change their gene expression. The mRNA profiles captured by TimeVault matched genetic responses measured using other methods, suggesting the recorder functions with high fidelity.

Another test showcased the time capsule’s power to observe complex diseases, such as lung cancer. Some tumor cells thwart medications and survive treatment. These cells don’t contain mutations that lead to drug resistance, suggesting they’re able to escape in other ways.

Using TimeVault, the team logged the cells’ activity before treatment began and discovered a ledger of genes, some previously not linked to cancer, that protect tumors from common therapies. By comparing gene expression from before and after treatment, they homed in on several overactive genes. Shutting these down boosted a cancer drug’s ability to kill more tumor cells, with one chemical cocktail lowering resistance to the cancer treatment.

The team is just beginning to explore TimeVault’s potential. One idea is to capture mRNA for longer periods of time from a single cell to record its unique genetic history. They’re also eager to re-engineer the technology so it works in mice, allowing scientists to capture an atlas of gene expression in living animals.

“By linking past and present cellular states, TimeVault provides a powerful tool for decoding how cells respond to stress, make fate decisions, and resist therapy,” wrote the team.

The post Scientists Turn Mysterious Cell ‘Vaults’ Into a Diary of Genetic Activity Through Time appeared first on SingularityHub.

2026-01-21 04:59:24

As data-center energy bills grow exponentially, technology companies are looking to nuclear for reliable, carbon-free power. Meta has now made an unusually direct bet on a startup developing small modular reactor technology by agreeing to finance the fuel for its first reactors.

The nuclear industry’s flagging fortunes have rebounded in recent years as companies like Google, Amazon, and Microsoft have signed long-term deals with providers and invested in startups developing next-generation reactors. US nuclear capacity is forecast to rise 63 percent in the coming decades thanks largely to data-center demand.

But Meta has gone a step further by prepaying for power from Oklo, a US startup building small modular reactors. Oklo will use the cash to procure nuclear fuel for a 1.2-gigawatt plant in Ohio that could come online as early as 2030.

The deal is part of Meta’s broader nuclear investment strategy. Other agreements include a partnership with utility company, Vistra, to extend and expand three existing reactors and one with Bill Gates-backed TerraPower to develop advanced small modular reactors. Together, the projects could deliver up to 6.6 gigawatts of nuclear power by 2035. And that’s on top of a deal last June with Constellation Energy to extend the life of its Illinois power station for a further 20 years.

“Our agreements with Vistra, TerraPower, Oklo, and Constellation make Meta one of the most significant corporate purchasers of nuclear energy in American history,” Joel Kaplan, Meta’s chief global affairs officer, said in a statement.

While utilities commonly negotiate long-term fuel contracts, this appears to be the first instance of a tech company purchasing the fuel that will generate the electricity it plans to buy, according to Koroush Shirvan, a researcher at MIT. “I’m trying to think of any other customers who provide fuel other than the US government,” Shirvan toldWired. “I can’t think of any.”

Part of the reason for the unusual deal is that securing fuel for advanced reactor designs like Oklo’s is not simple. The company requires a special kind of fuel called high-assay low-enriched uranium, or HALEU, which is roughly four times more enriched than traditional reactor fuel.

This more concentrated fuel is critical for building smaller, more efficient nuclear reactors. American companies are racing to grow the capacity to develop this fuel domestically, but at present, the only commercial vendors are Russia and China. And with a federal ban on certain uranium imports from Russia, the price of nuclear fuel has been rising rapidly.

Oklo will use the cash from Meta to secure fuel for the first phase of its Pike County power plant, which will supply the grid serving Meta’s data centers in the region. The facility is targeting a 2030 launch, though it won’t be producing the full 1.2 gigawatts until 2034.

It’s a somewhat risky bet for the tech giant. The Nuclear Regulatory Commission rejected Oklo’s licence application in 2022, and it has yet to resubmit. An anonymous former NRC official who dealt with the application recently told Bloomberg the company “is probably the worst applicant the NRC has ever had.”

But Meta isn’t putting all its eggs in one basket.

The deal with TerraPower will help fund development of two reactors capable of generating up to 690 megawatts by 2032, with rights for energy from up to six additional units by 2035. “We’re getting paid to start a project, which is really different,” TerraPower CEO Chris Levesque told The Wall Street Journal. “This is an order for real work to begin a megaproject.”

And the agreement with Vistra is more conventional. Meta is committing to purchase more than 2.1 gigawatts over 20 years from the existing capacity of the utility’s Perry and Davis-Besse plants in Ohio. It will purchase another 433 megawatts from expanding capacity at both plants as well as the Beaver Valley plant in Pennsylvania. All three plants had been expected to close just a few years ago, but Vistra is now planning to apply for licence extensions.

The three deals represent a bold bet on nuclear power’s potential to meet AI’s future energy demands. The big question is whether AI will still rely on the same kind of power-hungry models we have today by the time these plants come online next decade. Regardless, the current AI boom is helping power a nuclear renaissance that we may all benefit from in the years to come.

The post Meta Will Buy Startup’s Nuclear Fuel in Unusual Deal to Power AI Data Centers appeared first on SingularityHub.