In a sense, each of us begins life ready for action. Many animals perform amazing feats soon after they鈥檙e born. Spiders spin webs. Whales swim. But where do these innate abilities come from? Obviously, the brain plays a key role as it contains the trillions of neural connections needed to control complex behaviors. However, the genome has space for only a small fraction of that information. This paradox has stumped scientists for decades. Now, 黑料吃瓜资源 (CSHL) Professors Anthony Zador and Alexei Koulakov have devised a potential solution using artificial intelligence.
When Zador first encounters this problem, he puts a new spin on it. 鈥淲hat if the genome鈥檚 limited capacity is the very thing that makes us so smart?鈥� he wonders. 鈥淲hat if it鈥檚 a feature, not a bug?鈥� In other words, maybe we can act intelligently and learn quickly because the genome鈥檚 limits force us to adapt. This is a big, bold idea鈥攖ough to demonstrate. After all, we can鈥檛 stretch lab experiments across billions of years of evolution. That鈥檚 where the idea of the genomic bottleneck algorithm emerges.
In AI, generations don鈥檛 span decades. New models are born with the push of a button. Zador, Koulakov, and CSHL postdocs Divyansha Lachi and Sergey Shuvaev set out to develop a computer algorithm that folds heaps of data into a neat package鈥攎uch like our genome might compress the information needed to form functional brain circuits. They then test this algorithm against AI networks that undergo multiple training rounds. Amazingly, they find the new, untrained algorithm performs tasks like image recognition almost as effectively as state-of-the-art AI. Their algorithm even holds its own in video games like Space Invaders. It鈥檚 as if it innately understands how to play.
An AI-simulated cheetah cannot move forward on its own without training. Press play to see how it does with the genomic bottleneck algorithm.
Does this mean AI will soon replicate our natural abilities? 鈥淲e haven鈥檛 reached that level,鈥� says Koulakov. 鈥淭he brain鈥檚 cortical architecture can fit about 280 terabytes of information鈥�32 years of high-definition video. Our genomes accommodate about one hour. This implies a 400,000-fold compression technology cannot yet match.鈥�
Nevertheless, the algorithm allows for compression levels thus far unseen in AI. That feature could have impressive uses in tech. Shuvaev, the study鈥檚 lead author, explains:
鈥淔or example, if you wanted to run a large language model on a cell phone, one way [the algorithm] could be used is to unfold your model layer by layer on the hardware.鈥�
Such applications could mean more evolved AI with faster runtimes. And to think, it only took 3.5 billion years of evolution to get here.
Written by: Samuel Diamond, Senior Communications Strategist | [email protected] | 516-367-5055
Funding
Deep Valley Labs, The G. Harold & Leila Y. Mathers Foundation, Schmidt Futures
Citation
Shuvaev, S, et al., 鈥淓ncoding innate ability through a genomic bottleneck鈥�, Proceedings of the National Academy of Sciences, September 12, 2024. DOI: