This is not a fly loaded onto a computer

Last week, some buzz about the so-called virtual “embodied fly” in

The videos come from San Francisco-based Eon Systems, which says it’s working toward “digital human intelligence” and claims it wants to build a full digital simulation of a mouse brain within the next two years — an ambitious timeline, so to speak. Co-founder Alexander Wiesner Gross subscriber The original clip went public, calling it “the world’s first embodiment of a whole-brain simulation that produces multiple behaviors” and hinting at the impending technological singularity. CEO Michael Andrej to publish different pieces, Describe it As a “real animal uploaded”.

And that was for proof: no detailed methods, no scientific papers, no independent verification, just videos of what looks like a digital fly walking around, eating and rubbing its legs together.

related to artificial intelligence the accounts on X and Reddit Amplify Clips and repeat captions as fact. Expected approvals from leadership such as Elon Musk (“amazing”), Brian Johnson (“This is amazing”), and Peter Diamandis (“This is a living organism…on the Internet”) added fuel to the fire. Then content farms moved in, repackaging the whole thing as “news.” Celebration the Download Brain for the first time “And he asks”Are humans next?(Yes, they too indicated Matrix(And spoiler: we’re not next.)

The Internet was buzzing. The proof was still two short videos about X. If you’re going to tell the world that you’ve just achieved what could be one of the most important scientific milestones in human history, you better come with receipts.

Andregg tried to provide some clarity on X in a file string It was part set of warnings, part vaguely described scientific jargon, part concrete-sounding numbers like “Behavior accuracy 91%“I’ve sat with this scale for a while, and I still don’t really know what it’s supposed to mean, and I spent a good portion of my master’s degree studying animal behavior. And yet it is He insisted That “this is, in our view, a true free-raising animal.”

I sent Andregg a message on LinkedIn asking for more details. He responded with a link to a blog post Eon had just published published It’s titled “How the Eon team produced a virtual avatar fly.” It wasn’t a scientific paper, but it was something, I guess.

For experts Edge That I spoke to, that blog wasn’t nearly enough, but it took a much more careful line than the posts on X, in as much as not saying “this is a real fly.” While initial posts “withheld” important details about the work, the new blog provided more context, said Shahab Bakhtiari, a professor who leads the Systems Neuroscience and Artificial Intelligence Laboratory at the University of Montreal. He added: “But it arrived a little late, and is still not enough to fully verify the veracity of these allegations.” He was expecting a detailed technical report with details about things like software, code, and simulation environments that would allow other scientists to reproduce and evaluate the work.

Alexander Bates, a research fellow in neuroscience at Harvard Medical School who studies fly brains, echoed what Bakhtiari said. He said the group “didn’t accomplish anything” and that although the blog provided more details about what the team did — they stitched together existing large-scale projects such as a detailed map of the fly’s brain, a physical simulation of the fly’s body, and models simulating how these things would interact in a virtual environment — “for a claim of this size, I would expect something that should explain the whole approach in detail.”

Bates also said that the virtual fly’s behavior should be evaluated against real data and “clearly defined metrics,” adding that the 91% figure was not yet explained in the blog post. “Also, the fly does not fly.”

Bates said Edge He recognizes that “strong framing and hype can matter in fundraising,” but stressed that Eon’s claim of a “real animal uploaded” is not credible. Aran Naeby, a professor of machine learning at Carnegie Mellon University, said the group was “not even close” to capturing the entire fly brain, showing the connections between cells but not important details like neurotransmitters or how strong the connections are between different neurons. The engine system isn’t “true load” either, he said. “We’re not faithfully simulating his brain In silico“.

Well, let’s say Ion actually did that. He copied the fly’s brain perfectly. The whole thing. Every last bit. Do we have a digital fly now?

Yes. No. Maybe. I don’t know, and maybe you don’t either. Nor ion. The blog has conveniently ignored the important definitional questions at the heart of the upload claim: What exactly counts as a fly? When we think of flies, we don’t think of a set of behaviors or neural connections. We’re thinking, well, about the fly. Is it enough to reproduce some fly-like behavior in the simulation? Is the entire mapped brain computed in a virtual vessel? Or does “fly” mean the whole messy biological package – the body, the cells, the metabolism, and everything that might be considered “memory” or experience acquired over the course of its life?

This is the easy version of the problem. Obviously, the thing on the screen is not a fly at all. It is a composite of neural wiring, programming, and other information pieced together from many different animals. This is useful when modeling, but what object can we claim is loaded in this case? Technically, it’s also a copy, not a download, that comes with clear, deep effects that cut through the noise with ease: you can make two, 10, or 10,000 flies of “the same” fly. So what?

Normally, I wouldn’t expect a startup to solve a major metaphysical problem — philosophers have been bickering about this problem for centuries — but they’re the ones who say they have a “literal beast of burden.”

The experts I spoke to weren’t convinced the term even made sense. Bakhtiari said there was still an “open question” about whether it was possible for a “real animal to have been uploaded.” Jonathan Birch, a philosopher at the London School of Economics, was more blunt: “I don’t think we should ever say ‘loaded animal’.” What Ion aims to do is “mimicking the entire brain,” he said, leaving the rest of the animal behind.

Tom McClelland, a philosopher at Cambridge University, said biology is important for behavior. “So, at best, they uploaded some of the fly’s brain and thus uploaded some of the fly’s brain.”

Some time after his viral post, I asked Andrej if he stood by his claim. He told me: “Yes.” In fact, he went further: “We (the research group and academic collaborators) believe that this fly is conscious in a limited sense, that it can smell, see, taste, etc.” (I won’t go into the whole thing at all Awareness Thing.) He described the system as a kind of “MVP,” a minimum viable product, for a loaded animal, a product with “a lot of limitations.” I don’t quite know how to envision minimum viable flight. It’s a fly, not an app. MVP is the talk of technology startups, not science.

When I returned to Andrej a second time — this time after speaking with experts and relaying their criticisms — he still stood by the original claim, but with more caveats. He admitted that the work was “not an exact replica of the fly,” and added that Ion had never said that. “I don’t think of loading as a binary concept,” he told me, describing the “different levels” of loading and acknowledging that we don’t yet know how much biology is needed to capture important information. “There is a lot of work to be done to achieve the level of loading that we might want for ourselves one day.”