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.
Investor's Corner
New drone video shows Tesla’s Optimus Factory reaching a turning point
New drone footage shows Tesla’s dedicated Optimus factory steel frame nearing completion at Giga Texas.
Tesla’s dedicated Optimus factory at Gigafactory Texas is closing in on a finished steel frame, according to drone footage posted Thursday afternoon by longtime site observer Joe Tegtmeyer. In the video, Tegtmeyer said structural steel assembly is now about five column grids away from reaching the building’s north perimeter beam, putting the primary skeleton in its final stretch roughly six months after Tesla broke ground on the North Campus site in late March.
The Giga Texas Optimus Factory latest update … construction keeps moving fast!
Steel assembly is now only about 5 column grids from the north perimeter beam. Concrete is going in on the three upper floors, rebar is still being laid for more pours, and footing / grade-beam work… pic.twitter.com/qvFPvwnYzW
— Joe Tegtmeyer 🚀 🤠🛸😎 (@JoeTegtmeyer) September 17, 2026
Tegtmeyer’s footage shows concrete already going in on three upper floors while crews continue laying rebar and pouring grade beam footings at ground level. That kind of parallel work, steel rising at one end of the site while concrete sets at the other, is a scheduling approach Tesla used at the original Giga Texas building and appears to be repeating here to save time before the plant’s targeted 2027 production start.
Teslarati has tracked the building’s progress since Tesla confirmed construction was officially underway in May, when the first steel structure went up on what was then bare, reclaimed land. The facility is part of a more than 5.2 million square foot expansion of Giga Texas’s North Campus that Tesla has said will eventually run nearly the length of the existing vehicle factory, over 4,000 feet, while sitting somewhat narrower. Musk has pegged the long term output target at 10 million Optimus units a year once the line is running at full capacity, a volume that would dwarf the one million unit pilot line Tesla is standing up separately at its Fremont, California factory.
Tesla Giga Texas to feature massive Optimus V4 production line
The Texas facility sits alongside another major buildout on the same campus. Terafab, the joint Tesla and SpaceX chip fabrication plant that will eventually supply the silicon running Optimus units in the field. Housing robot assembly and chip production on the same grounds is a deliberate supply chain decision, cutting down on the shipping and lead time that would otherwise sit between the two.
Tesla has not given an updated timeline beyond its previously stated goal of bringing high volume Optimus production online at the site in the summer of 2027. Fremont’s smaller pilot line began mass producing the current Gen 3 robot in January, with that plant expected to build tens of thousands of units this year primarily to generate the real world data Tesla needs to refine the robot’s software before Giga Texas ramps up. Six months of visible construction progress, tracked almost entirely through Tegtmeyer’s recurring drone flights, gives the clearest outside look yet at how seriously Tesla is treating that 2027 deadline.
Investor's Corner
Tesla and SpaceX take “Terafab” Trademark fight to Federal Court
Tesla and SpaceX sue a small Illinois firm after cease and desist letters over Terafab.
Tesla and SpaceX are asking a federal judge to rule that their planned Terafab chip factory does not infringe a small Illinois company’s trademark, a request that arrives only after months of quiet negotiation broke down this summer.
The dispute traces to May 18, when Tesla filed three U.S. trademark applications for “Terafab” and “Tesla Terafab,” covering semiconductor chips and related chip making services. TERA-print LLC, a nanotechnology company that has held a federal trademark for “Tera-Fab” since 2021, responded five days later with a cease and desist letter. According to the lawsuit, first reported by Reuters, TERA-print argued that Tesla and SpaceX’s use of “Terafab” would confuse consumers familiar with its own trademark, which covers a desktop photolithography printer sold to researchers for sensor and bioengineering work.
What stands out in the filing is the timing of TERA-print’s own paperwork. One day before sending that cease and desist letter, on May 22, TERA-print applied to expand its existing registration to cover semiconductor materials, silicon chips, nanoelectronic devices and AI design services, categories it had not previously claimed. Tesla and SpaceX call that filing opportunistic in their complaint, noting it arrived two months after Tesla’s public Terafab announcement and just days after Tesla’s own trademark applications went in.
Elon Musk launches TERAFAB: The $25B Tesla-SpaceXAI chip factory that will rewire the AI industry
By June 10, TERA-print was threatening to sue for federal trademark infringement, false designation of origin and unfair competition, the complaint states. Rather than wait to be sued, Tesla, SpaceX and SpaceXAI met with TERA-print six separate times between June and August trying to resolve the dispute directly. Those talks collapsed, and the companies filed for declaratory judgment this week in the U.S. District Court for the Western District of Texas, asking a judge to find that “Terafab” does not infringe TERA-print’s mark before TERA-print can file a claim of its own.
TERA-print isn’t backing down. The company told PCMag it discussed a settlement with Tesla as recently as September 2 and feels misled by what it called Tesla’s professed interest in settling. Its CTO, Andrey Ivankin, said TERA-print holds a Defense Department contract to fabricate semiconductors and partially owns Mattiq Inc., an AI company built on TERA-print’s products, and that the company will vigorously defend its rights.
Tesla and SpaceX argue the overlap is superficial. Terafab is planned as a $16.8 billion complex spanning roughly 100 million square feet at the Grimes County site SpaceX confirmed last month, built to produce chips for Optimus robots, Tesla’s AI computing needs and SpaceX’s orbital data center ambitions, a scale and purpose the companies say no reasonable consumer would confuse with a tabletop lab printer. TERA-print’s product line has stayed focused on lithography tools for biological and sensor research since it registered its mark in 2021.
The trademark fight is the second legal dispute tied to the Terafab project in the past week, following a separate SpaceX suit aimed at keeping company records about the facility out of public view, as KBTX reported. Whether construction proceeds under the Terafab name now depends on a federal judge in Austin.
News
NHTSA just escalated its Tesla Cybercab investigation in a big way
NHTSA escalated its Cybercab audit into a sworn Special Order with a September 30 deadline.
Federal regulators have moved from asking Tesla questions about its Cybercab to demanding sworn answers. The National Highway Traffic Safety Administration issued a Special Order that requires a Tesla officer to sign an affidavit attesting to the completeness of the company’s responses, with a deadline of September 30.
The order builds on Audit Query AQ26002, which NHTSA opened on September 3, the same day Tesla began commercial Cybercab service in Austin. Teslarati covered that initial inquiry when it surfaced, noting the agency wanted to understand how Tesla certified a vehicle with no permanently attached steering wheel, pedals, or mirrors as compliant with Federal Motor Vehicle Safety Standards. A Special Order is a different tool and converts a fact finding review into a legally enforceable demand, the same mechanism NHTSA used against Tesla in 2023 during its Autopilot investigation.
Several of the 21 requests target a specific gap in Cybercab’s design. One asks whether Tesla used temporarily attached human controls at any point to help certify the vehicle, and if so, which standards depended on that equipment being present. Another quotes an existing rule directly: “The service brakes shall be activated by means of a foot control.” Cybercab has no foot pedal. NHTSA wants a detailed explanation of how the vehicle satisfies that requirement, and how it complies without the kind of exemption granted to Zoox in July under Part 555, the regulatory pathway built for steering wheel free vehicles.
The order does not claim Cybercab is unsafe or that Tesla broke a rule. It requires Tesla to explain, under oath, the reasoning behind decisions the company already made when it self-certified the vehicle. That distinction matters, but so does the exposure. Motor1’s reporting, summarized here, put potential civil penalty exposure as high as $139 million if NHTSA later finds the certification was flawed, on top of whatever criminal risk comes with a false sworn statement.
Tesla has not said publicly how it plans to respond. Cybercab is still carrying passengers in Austin through the Robotaxi app while the September 30 deadline approaches, and the company has continued expanding the vehicle’s footprint even as the regulatory question remains open. The Special Order does not pause any of that and just sets a date by which Tesla has to put its certification logic on the record, with a company officer’s name attached to it.
New drone video shows Tesla’s Optimus Factory reaching a turning point
Tesla and SpaceX take “Terafab” Trademark fight to Federal Court
