News
Scientists use AI neural network to translate speech from brain activity
Three recently published studies focused on using artificial intelligence (AI) neural networks to generate audio output from brain signals have shown promising results, namely by producing identifiable sounds up to 80% of the time. Participants in the studies first had their brain signals measured while they were either reading aloud or listening to specific words. All the data was then given to a neural network to “learn” how to interpret brain signals after which the final sounds were reconstructed for listeners to identify. These results represent hopeful prospects for the field of brain-computer interfaces (BCIs), where thought-based communication is quickly moving from the realm of science fiction to reality.
The idea of connecting human brains to computers is far from new. In fact, several relevant milestones have been made in recent years including enabling paralyzed individuals to operate tablet computers with their brain waves. Elon Musk has also famously brought attention to the field with Neuralink, his BCI company that essentially hopes to merge human consciousness with the power of the Internet. As brain-computer interface technology expands and develops new ways to foster communication between brains and machines, studies like these, originally highlighted by Science Magazine, will continue demonstrating the steady march of progress.

In the first study conducted by researchers from Columbia University and Hofstra Northwell School of Medicine, both in New York, five epileptic participants had the brain signals from their auditory cortexes recorded as they listened to stories and numbers being read to them. The signal data was provided to a neural network for analysis which then reconstructed audio files that were accurately identified by participating listeners 75% of the time.
In the second study conducted by a team from the University of Bremen (Germany), Maastricht University (Netherlands), Northwestern University (Illinois), and Virginia Commonwealth University (Virginia), brain signal data was gathered from six patients’ speech planning and motor areas while undergoing tumor surgeries. Each patient read specific words aloud to target the data collected. After the brain data and audio data were given to their neural network for training, the program was given brain signals not included in the training set to recreate audio, the result producing words that were recognizable 40% of the time.
Finally, in a third study by a team at the University of California, San Francisco, three participants with epilepsy read text aloud while brain activity was captured from the speech and motor areas of their brains. The audio generated from their neural network’s analysis of the signal readings was presented to a group of 166 people who were asked to identify the sentences from a multiple choice test – some sentences were identified with 80% accuracy.
While the research presented in these studies shows serious progress towards connecting human brains to computers, there are still a few significant hurdles. For one, the way neuron signal patterns in the brain translate into sounds varies from person to person, so neural networks must be trained on each individual person. The best results require the best data possible, i.e., the most precise neuron signals possible, meaning this is something that can only be obtained by placing electrodes in the brain itself. The opportunities to collect data at this invasive level for research are limited, relying on voluntary participation and approval of experiments.
All three of the studies highlighted demonstrated an ability to reconstruct speech based on neural data in some significant capacity; however, also in all cases, the study participants were able to create audible sounds to use with the computer training set. In the case of patients unable to speak, the level of difficultly in interpreting the brain’s speech signals from other signals will be the biggest challenge. Also, the differences between brain signals during actual speech vs. thinking about speech will complicate matters further.
Elon Musk
Sean Duffy fires back at Elon Musk over free Starlink
Sean Duffy says the FAA picked a ground network over Starlink, and Musk isn’t impressed.
Transportation Secretary Sean Duffy has responded to Elon Musk’s claim that the Federal Aviation Administration turned down free Starlink terminals before last week’s nationwide ground stop, saying the department chose a ground-based network after running into problems with an earlier version of the satellite service. Musk’s reply was a single emoji: “🤦♂️.”
Speaking at a press conference, Duffy said Starlink had been useful in some parts of the system. “I’ve worked well with Elon, and he, Starlink was a great solution for some of our assets in Alaska, but we sent Starlink, in one of its prior iterations, to our tech lab, and, there was some issues that we had to navigate, and we thought that the best method is to have a network that’s based on the ground,” Duffy said, according to a transcript published by ZeroHedge. “Now, I think what he’s doing with Starlink and the speed is remarkable, but, our airspace is very big. We have assets everywhere. And so, again, we used him for different parts of the airspace.”
Duffy did not dispute that the offer was made. He also did not say what problems the FAA found during testing.
The exchange started Monday, when Musk wrote on X, “Free Starlinks for FAA were offered to Sec Duffy, but he said they were not needed. As it turns out, they were,” as Teslarati reported. That post came three days after the FAA grounded departures nationwide on November 13, when its legacy equipment became unstable as data flooded its telecommunications network at 1,000 connection attempts every four to five seconds. The FAA traced the flood to a radar site in Amarillo, Texas, and a radio site in El Dorado, Arkansas, where L3Harris had been installing a new router. Roughly 12,462 flights were canceled that day.
As I was saying …
Free Starlinks for FAA were offered to Sec Duffy, but he said they were not needed. As it turns out, they were.
— Elon Musk (@elonmusk) November 16, 2026
Duffy has also signaled frustration with L3Harris, which runs the FAA’s current network. “I want to get away from L3Harris and get to a new one, but they own the network,” he said. “I can’t cancel the network that they own.” He has asked Congress for $50 billion to modernize air traffic control, arguing against “a patchwork, a fix to a system that’s critical for every single family.”
The FAA’s position on satellite links predates this week. In a July 2025 letter to lawmakers, FAA Administrator Bryan Bedford said the agency “would not use Starlink (or any satellite service) as the sole communications technology” for safety-critical air traffic services. Reuters reported at the time that the FAA was not considering replacing its L3Harris and Verizon contracts with Starlink. Verizon’s FAA Enterprise Network Services contract, awarded in 2023, is valued at up to $2.4 billion over 15 years.
Musk first pressed the issue in February 2025, while he was leading the Department of Government Efficiency, writing that “the Verizon system is not working and so is putting air travelers at serious risk” before correcting himself days later to blame L3Harris. SpaceX said at the time it was supplying Starlink kits to the FAA at no cost and had no intent to take over any existing contract.
Elon Musk
Tesla gears up Cybercab with Starlink for public debut, but with notable changes
Tesla is gearing up Cybercab, its all-electric and fully autonomous ride-hailing vehicle, for a very public debut as it plans to start customer deliveries in December. However, the units that rolled out of Gigafactory Texas this morning seem to differ from what we have seen previously.
Aerial images captured by drone photographer Joe Tegtmeyer at Giga Texas on November 18 reveal a noticeable update to the appearance of Tesla’s Cybercabs equipped with Starlink. Rows of gold production vehicles sit in the outbound lot, each carrying a flush-mounted Starlink V5 terminal on the rear hatch that looks more refined than the version first shown months earlier.
When Tesla first unveiled the integration in July and displayed a physical example in August, the Starlink hardware appeared as a relatively distinct rectangular panel on the hatch. Later September footage of large numbers of Cybercabs leaving the factory still showed a lighter, more obvious rectangular cover that roughly matched the vehicle’s gold paint:

Images via Joe Tegtmeyer
The newest batch stands out for how cleanly the cover blends into the roof’s character lines. Observers and replies to Tegtmeyer’s post describe the result as tighter and more intentional, sitting flush and following the body contours rather than reading as an added panel.
All of the Cybercabs in the latest images also wear prominent “Cybercab” logos on the doors and rear hatch, confirming these are finished production units rather than early prototypes. They are staged near the factory building, end-of-line areas, and covered transport trailers, with some already lined up for shipment.
The scale and consistency of the installation point to factory tooling rather than retrofit work.
A secondary detail noted in related sightings is a darker tone to the Starlink cover compared with earlier lighter panels. Tegtmeyer himself left open whether the darker finish is deliberate. There are a variety of explanations for it: perhaps for thermal performance, signal characteristics, or simple visual identification—or simply a batch variation still being refined.
Either way, the overall effect is a more integrated, production-ready appearance. For a Robotaxi that relies on Starlink for fleet connectivity, passenger services, and backup data links in areas with weak cellular coverage, the cleaner hatch design is more than cosmetic.
It signals that Tesla is polishing the final visual and hardware details as Cybercab volumes continue to climb at Giga Texas ahead of broader Robotaxi deployment. The vehicles now look to be built with Starlink instead of it being added afterward.
Elon Musk
Tesla’s Powercore could change how data centers get power
Tesla unveiled Powercore, a solid-state transformer that could ease one of the grid’s worst bottlenecks.
Tesla has unveiled Powercore, a solid-state medium voltage transformer that also handles the jobs of an inverter and a circuit breaker in a single unit. “Solid state medium voltage transformer, inverter and breaker – reliable, safe and secure,” the company’s Tesla Megapack account posted on X Tuesday, adding that the product is “Made in America” and will be “Operating 2027.” Elon Musk followed with his own post: “Advanced solid state transformers at scale from Tesla.”
Tesla rates Powercore at 7.4 MW, or 9 MVA, of continuous output at 34.5 kV AC, with grid-forming support for 10 MVA of peak power. Specs shared by Sawyer Merritt list a peak efficiency of 98.86 percent, a DC range of 800 to 2,000 volts and a built-in solid-state breaker that clears faults in about 100 microseconds. Tesla has not announced pricing, a factory location or a first customer.
The product fills a gap Tesla left open last year. When it introduced Megapack 3 and Megablock in September 2025, Tesla said it would build its own transformers, but shared almost nothing about them. Megablock groups four Megapack 3 units with a transformer and switchgear into a single medium voltage block, as Teslarati reported. Powercore appears to be that missing piece.
A typical battery or solar project today uses low voltage inverters to turn DC into AC, then a large oil-filled transformer to step that power up to grid voltage. A solid-state unit handles both steps with power electronics, which means less equipment, a smaller footprint and no flammable oil. Grid-forming capability lets it actively help hold voltage and frequency steady rather than simply following the grid.
Introducing Powercore Solid state medium voltage transformer, inverter and breaker – reliable, safe and secure 7.4 MW/9 MVA continuously at 34.5 kVac w/ grid-forming support for 10MVA of peak power Made in America. Operating 2027 https://t.co/sHxqnpWJrf — Tesla Megapack (@Tesla_Megapack) November 17, 2026
Timing is a big part of the pitch. Lead times for conventional transformers have stretched into years as AI data centers and renewable projects crowd utility interconnection queues. Tesla energy executive Max de Zegher pointed to the problem in a post Musk shared, noting that “transformers have barely changed since the 1890s: copper coils in a tank of oil,” while batteries, solar and compute all run on DC. Powercore, he wrote, delivers “one conversion and about half the losses.” Tesla has also said its Megapack supply for 2027 is nearly sold out.
Tesla deployed 13.7 GWh of energy storage in the third quarter, its second best quarter on record. On Tuesday, the company also said customer solar systems deployed by Tesla Energy now generate more than twice the electricity its global operations consume each year.