Episode 3 of the Tech & Business Radio Show on WAMH Radio at Amherst College.
What is up and welcome to the March 12th. episode of Tech Business Radio. You're tuned in. Seven to eight. an hour of what's actually happening in tech. AI and business. just the real developments that matter. I'm pierce, and tonight to get into guys. We're talking about the first ever brain upload. a fly brain. running inside of a computer. for covering the anthropic family. what it actually means for future of AI and the federal government. and we'll also break down why open AI is leadership. is rating $25 million checks. to Donald Trump's super PACs. We've got some big announcements. an interview drop in Thursday. and a returning guest. that you would not ever.
techandbusiness.org. all of our articles written by young professionals who actually work in these fields. also on LinkedIn. and our podcast, as well as the most recent episode. is on Spotify. So essentially, you guys have to stay curious. because that's not cliche.
On March 7, a company called Eon Systems out of San Francisco released a video of a fruit fly, walking, grooming, and feeding inside of just a computer simulation. But here's the thing. It wasn't even animated. It wasn't a model. It wasn't programmed to do any of these things. They took the complete wiring diagram called the connectome of a real fly brain, over 125,000 neurons, 50 million synaptic connections.
Here's where it connects to everything that we covered. There's two black boxes that the tech work... neurological black box which is understanding how biological brains actually produce behavior and the AI black box which is understanding how LLMs reason. They're not separate problems. LLMs are at their core simulations of neurological pattern recognition. However, transformer architectures process information in ways that researchers are now finding. that genuinely mirror how certain brain regions handle language and spatial reasoning. Meadows researchers just published some work on opening the LLM black box, using something called circuit-based reasoning verification. where they trace the actual computational graphs inside the models to find where reasoning breaks down. at the same time though. Eon is showing that if you faithfully replicate the biological wiring. The behavior just emerges without training. The dual mandate is this, the better we understand the neurological black box, the better we can inform the AI of black box and vice versa. These fields are converging. and the fly upload is just the proof of concept. Eons next target is the mouse brain with 70 million neurons, 560 times the fly. human scale emulation. or watching like the Wright Brothers moment of brain emulation and most people. but this is huge and literally something out of a sci-fi novel.
Pentagon, they wanted anthropic to agree that its AI could be used for AI lawful purposes. They wanted explicit language banning two things, domestic mass surveillance and fully autonomous weapons. The Department of Defense agreed to almost every term except for moving the prohibition on analysis of bulk acquired data. Anthropic wouldn't budge on that in the deadline path. Trump went on truth social, calling Anthropic a radical left woke company, and ordered every federal agency to stop using their tech. Hours later, the Pentagon formally designated anthropic a supply chain risk to national security.
wrote a 1,600-word internal memo to staff that leaked through the information. He called OpenAI's Pentagon deal Safety Theater, called Sam Altman's public statements straight-up lies, said the reason OpenAI accepted the deal was that they cared about placating employees, while Anthropic actually cared about preventing abuses. He also wrote that the administration dislikes Anthropic because they haven't donated or offered dictator-style praise to Trump. The CEO then apologized for the tone, saying it was written in the heat of a chaotic day.
As Congressman Sam Liccardo pointed out, the problem with relying on existing law is that there is no law. The law is always years behind technology. This isn't happening in a vacuum. The Pentagon has renamed the Department of War under this administration. Pete Hegseth is secretary, hours after this whole.
donated $25 million to the MAGA Inc. Super PAC. Sam Altman donated $1 million to Trump's Inaugular Fund and was at the White House for the Stargate announcement. The Quit GPT Boycott campaign hit over 700,000 supporters. Ice is reportedly using chat.
Competition and market forces need to work in tandem with vigilant oversight, just like we wrote about with net neutrality, when one player aligns this closely
It's an industrial organization story at its core. Just like the IBM Red Hat piece showed how vertical integration reshapes market structure. The anthropic Pentagon open AI triangle is about who gets vertically integrated into the national security apparatus.
launched with reportedly 1 trillion parameters, but only 32 billion active per- Alibaba's Quen 3.5 Small series has a 9 billion parameter model. batching output from models 13 times its size.
We literally cannot match this because of the fact that we have too many energy regulations, which is good.
Apple's Siri is now being powered by Google's 1.2 trillion parameter Gemini model, running on Apple's private cloud compute. Compute infrastructure race is just as important as the model race. In terms of the big picture, that was a lot of numbers, a lot of companies, but every major release in 2026 emphasized...
the difference between a chatbot and an employee and that's where the business implication
Protocol Labs taught neurons to play pawns on a system called dish brain. That took 18 months. Then they moved to Dune, a 3D environment with enemies, navigation, spatial decisions. And the neurons learned it in one week on their new CL-1 Biologer. rented a CO1 through their cloud API and wired it into a small The neurons are nudging the model's token selection in real time. The brain cells aren't writing poetry, but they're sitting inside an AI stack and modulating output.
What people are talking about right now is efficiency and the implications on the environment with the energy demands.
Everyone's favorite big finance bank executive said last month that now is the time to start planning for large scale AI labor displacement during the same time as record lows in employment for college educated.
These products are entering clinical environments without sufficient testing on actual patients. The FDA has approved over 1,300 AM medical devices, and submissions are spiking, but agentic AI systems that act autonomously improve themselves don't fit neatly into the existing regulatory framework. business story under every other story we covered. When we wrote about becoming AI first. When Joey wrote about the shift in investment banking culture. When Leo covered business automations. All of it points here. The agentic capabilities in every model released this month are in abstract. They're tools that let one person do the work of five. That's either an opportunity or a threat, depending on which side of the disruption you're on. Our job in tech and business is to make sure you're on the right side of it, and that's why why.
to the millions listening at home. soon this time. for some of the largest media announcements. in the history of media.
