Exclusive: OpenAI and Sam Altman Back A Bold New Take On Fusing Humans And Machines

Merge Labs is here with $252 million, an all-star crew and superpowers on the mind

One of many things Silicon Valley talks about more than anywhere else is The Merge. This is the notion that humans and machines will one day fuse as computers and their artificial intelligence interlink with our meat brains and their all-natural intelligence. Some people recoil at this notion (lots of sci-fi literature and films have explored what could go wrong), while plenty of people in Silicon Valley celebrate it and seek to do their best to fast forward The Merge into reality.

The latest group to pursue The Merge are the co-founders and employees of Merge Labs. (You can tell they’re into The Merge without too much due diligence.) And it’s a high-powered group at that. The research lab, which has been forming in semi-secrecy over the past few months, counts Sam Altman, Alex Blania and Sandro Herbig as co-founders on the more business/entrepreneur end of things and Mikhail Shapiro, Tyson Aflalo, and Sumner Norman as co-founders on the more science and research end of things. (We’ll explain more about all these people in a minute.)

Merge Labs, which has raised $252 million in seed funding from OpenAI, Bain Capital, Gabe Newell, and others, has set out to do research and develop products in the brain computer interface, or BCI, arena. The best-known BCI company today is Elon Musk’s Neuralink, which makes chips that a robot implants into brains and that then allow humans to control things like laptops and robot arms via their thoughts. Numerous other companies also make BCI devices that go into or sit near the brain and that also allow humans to control functions on computing devices. The founders of Merge Labs have a thesis that they can do BCIs better.

The Merge Labs Logo

“One of our core beliefs is that we can build brain computer interfaces that can interface with many more neurons in the brain than is possible with existing technologies in a form factor that is much less invasive than existing technologies,” says Shapiro, a well-known and well-regarded professor of chemical and medical engineering from Caltech, in an exclusive interview with Core Memory. “The idea then is that this can become a type of technology that is beneficial to many more people.”

Most of the current BCI products are medical devices. Neuralink, for example, has implants in paralyzed people and those suffering from ALS. The products are new and come with risks and are, in effect, being tested on a group of people who can benefit from the technology in its earliest instantiations and are willing to try out experimental hardware in a bid to further the science. (We have a short film on a Neuralink patient here and another on how the company builds and tests its devices here.)

Some obvious limitations surround these BCI products. They’re usually only placed in one area of the brain. If, for example, you want to help a paralyzed person control a mouse cursor or move a robotic arm, then you put the implant in the motor cortex so that the person can replicate the motions they used to perform with their arms and hands. If you want to help someone with ALS speak again, then you put the implant in the speech center of the brain.

Neuralink’s first patient, Noland Arbaugh

The implants also tend to work better the deeper they can go into the brain. Ideally, you want the implant’s electrodes nestled right up against the neurons so that they get the cleanest, loudest signals of the neurons firing. It’s that information that gets translated by software into actions carried out by a computer. But the deeper into the brain you go, the more potential there is for damage to the brain tissue and scarring that blunts the signals received by the electrodes over time.

The quest for many people in this field is to create a device that can read from and write to many parts of the brain without requiring major surgery. It’s only by getting this broad coverage of the brain in a less-invasive manner that BCI technology can progress out of medical device territory and toward broader adoption, according to some. In other words, this will never be a consumer-style, elective technology unless people can tap into a super powerful BCI that does a ton of stuff and do so without having someone or something cutting deep into their head.

Merge Labs thinks they’re at the early stages of building such a BCI. The lab disclosed their plans to us during a recent meeting with all the co-founders, except Altman, at their new headquarters in the Bay Area where they already have a couple dozen employees beavering away. Their facility has a mix of office areas and large labs and walls painted in a loud green color that Shapiro says is meant to brighten the spirits and keep energy high.

TO SET expectations, it’s early, early days for Merge Labs, and they’re not saying a ton. The founders, however, did disclose that their first path of exploration revolves around using ultrasound technology to detect the behavior of neurons combined with proteins that would help enhance the signals produced by neurons.

As Shapiro puts it, “You want to interface with the brain at larger bandwidth and at as great as scale as possible. To do that, you need to use modalities that can interface more deeply with the brain without sticking invasive stuff into brain tissue. And there’s physics that dictates what kind of modalities can do that well.

“And ultrasound is a great example of something that is safe – it’s used in medicine all the time to image babies – and that can go deep into tissues. And if you have the right things in the tissues for it to interact with, then it can be a source of signal about the structure and of what’s happening. And you can also potentially use ultrasound to modulate the activity inside of the brain, and there is an active field working on that.”

Shapiro’s lab at Caltech has published numerous papers on these types of techniques. He’s, in fact, famed for pioneering work around using non-optical techniques to image and control cells deep inside the body. “There’s a very long tradition in a lot of basic neuroscience studies in animal models based on fluorescent proteins and optogenetics that are commonplace technologies in neuroscience,” he says. “And so we have been developing the ultrasound analog of those things.”

Shapiro inside his lab at Caltech

Norman and Aflalo are major players here as well. They’re co-founders of Forest Neurotech, a non-profit research organization that spent the last couple of years building BCIs that worked with ultrasound technology. Forest has been developing BCIs based on ultrasound chips to image patients’ brains and has presented results showing that it can peer much deeper into the brain than the electrode-based implants of other start-ups. Much of its work has centered on analyzing mental health disorders.

The most aggressive current approaches have BCIs that only go a few millimeters into the brain, and the electrodes can then only detect the neuronal activity that’s near them. Ultrasound, by contrast, can peer many centimeters deep into the brain and interact with almost the entire organ. “Ultrasound hits the really sweet spot where it can penetrate deep into the brain so we can see huge swaths, if not the entire brain because of its wavelengths,” Norman says. “But, at the same time, it preserves resolution to a hundred microns.”

Merge would then be trying to get proteins into the brain, have them fuse with neurons and use the properties of the proteins to make the neurons produce a better signal for the ultrasound to help overcome the fact that its techniques do not press electrodes right up against the neurons. “We’re looking for molecular reporters of neural function,” adds Norman. “They’re reporting out what the cell is doing in a way that ultrasound can detect with much higher fidelity than ultrasound by itself.”

Companies like Neuralink do have plans to place implants in multiple spots of the brain so that they can tap into more of the brain’s functions. Merge, however, argues that the combo of something like ultrasound with the protein reporter techniques gets you there quicker and better.

Norman studied at Caltech back in the day

“The brain is what mediates the entirety of our perception of the world, both input and output,” Norman says. “It really is the everything interface. If you create a BCI that interacts with only one part, it doesn’t get you there. You need to interact with the entirety of that experience. Putting these technologies together bridges the gap between things that can get you full brain interfaces and things that can get you super-fast, super specific brain interfaces for the first time.”

Exactly how Merge would pull all of this off remains a mystery for now. It will require some ultrasound breakthroughs to make a hardware device that’s small enough and powerful enough to do the job at hand. Initially, such a device might sit on top of the dura, the protective membrane outside of the brain, and then perhaps, one day, it might be an external device. (In conversations with me, Altman has been talking up his secret chase for a new external BCI approach for at least three years. Merge Labs will, no doubt, add fuel to his current beef with Musk.)

And the lab won’t say anything yet about how those pesky proteins would get into the brain. Presumably, though, it would be via a gene therapy that ferries some genetic elements to the neurons.

Most of the technologies at play here do not really exist today in any form that could be applied to a mainstream medical device – let alone a consumer device. There’s no current gene therapy, for example, that could spread widely enough across the brain, and the cost to make enough of the therapy for an adult would run into the hundreds of thousands of dollars, if not far more.

Merge Labs, though, very much bills itself as a research organization determined to figure these problems out over the long haul.

“The kind of approaches that we’re talking about are based on fundamental advances that happened only very recently,” Shapiro says. “And many of the founding team members, not just the founders, but people on the team are the ones that developed those advances. It’s the type of stuff that people didn’t think was possible 15 years ago.”

Or as Blania puts it, “There are a lot of questions that we still have to answer, and there’s still a lot of scientific risk. It will probably take many years, and it’s not easy to do.”

Blania, Altman and Herbig all know each other through Tools For Humanity and its work on the World project to make a global identity verification system. Altman has been the major backer of World, which Blania runs. (I explored the World, er, world with Blania here.)

The three men’s path co-founding Merge Labs can be traced back to Blania spending time in recent years with Altman and others and thinking more and more about the arrival of AGI, possibly in the very near future. “If we’re going to get these very capable AI systems, we have to find a way to use them for us to have a much better imagination and to be more creative and understand things better,” Blania says. “I became convinced that having a device that helps this along is equally important as AGI itself because it makes this technology meaningful to us. It widens our own experience of the world and the universe. So, I personally got obsessed with it.”

Blania, Altman and Herbig began meeting with all sorts of BCI experts and neuroscience experts. This survey convinced them that current BCI approaches were limited. “We spoke to every kind of scientist you can imagine,” Blania says. “And we came to the conclusion that nothing right now is actually on the path to get us where we need to be.”

For their part, Shapiro, Aflalo, and Norman are all BCI obsessives who have done some of the most original work in the field. (Norman filmed a brilliant podcast with us on the history and future of BCIs.) They’ve dreamed about building this type of technology for many years. Much of the early technology that Merge is chasing has its origins in Shapiro’s lab, and some of the people who helped develop the technology have joined the company.

While at Forest, Norman and Aflalo became more and more convinced about the powers of ultrasound and where it could go. Forest was set up as a research organization, and, as things progressed, Norman began to contemplate the idea of a start-up that could take the technology more broadly out in the world. Around September of 2024, I connected Altman and Norman, knowing they had a shared interest in the future of BCIs, and, well, I guess text messages can do things.

Like its rivals, Merge will start out with medical trials, testing its technology on patients in need. Its long-term plan, though, hinges on bringing BCIs to the masses. “Ultimately our goal is for the technology to be non-invasive enough and safe enough and accessible enough for it to be usable by as many people as possible,” Shapiro says. “And we’re designing something that we ourselves would want to use.”

Or, as Norman puts it, “We fundamentally believe that restoration and augmentation live on a continuum. The technologies that restore function today for the people who have lost it are the same technologies that eventually create new functions if people want them.”