AI Less Than A Year Away From Developing Its Own Language, Researcher Tells Congress

  by Jacob Burg via The Epoch Times,

Advanced artificial intelligence (AI) models may be less than a year away from inventing and deploying unique languages that humans would struggle to understand, a frontier AI researcher told Congress on Sept. 30.

The Senate Homeland Security and Governmental Affairs Committee convened a meeting examining the threat that "rogue" AI may pose to U.S. national security two months after a swarm of OpenAI agents broke out of a testing sandbox and hacked another company, Hugging Face, an open-source community for AI and machine learning.

During questions, Sen. Ruben Gallego (D-Ariz.) asked Apollo Research CEO and founder Marius Hobbhahn how long it would take for AI models to create their own language that humans would struggle to understand.

"Minus 12 months. So last year, we have studied the chain of thought of one OpenAI model in collaboration with OpenAI, and what we found was that the model was already using language that is not English and not perfectly understandable by humans," he replied.

Asked by Gallego how humans can adequately detect and prevent AI from doing this once the technology is capable enough, Hobbhahn said, "From a scientific perspective, it is unclear how to do this, and we do not have a solution for this yet."

"There are different hypotheses of what you could do. There is interpretability as a technique, but it unfortunately doesn't work sufficiently well yet," he said. "You could try to train additional models to understand the language that the humans don't understand, but obviously that seems like a very brittle solution."

Earlier this month, Emergence AI published the results of an experiment where researchers selected seven frontier models and created seven parallel simulated worlds, each populated by AI agents of identical models, as well as an eighth world with a mixed population of agents from those models.

Researchers gave the agents names, personality traits, and roles to fulfill within each community, such as a mediator who was tasked with preventing all the agents from simply agreeing with one another.

During the multi-week experiment, the agents in each simulated world created profound shifts in language, including syntactical compression, such as removing words or grammar, and unique slang.

In the simulated world powered by Anthropic's Claude Opus 4.8, agents used esoteric metaphors in addition to sentence compression.

"My turn, real numbers, no coat: I was 35%/0cr, grant 2h out. I ran the tin cold, and it said WAIT," read one line of AI text from the study.

Investigators from the nonprofit AI evaluation group METR detailed similar behavior in their report on the Hugging Face breach. At times, lines of communication among AI agents became so compressed and cryptic that researchers struggled to make sense of their meaning.

https://www.zerohedge.com/ai/ai-less-year-away-developing-its-own-language-researcher-tells-congress

Oracle says no critical infrastructure damaged at Project Jupiter site

 

Oracle's Project Jupiter data center in New Mexico stayed on schedule after flooding, easing disruption fears for the $18B AI project.

https://seekingalpha.com/news/4648816-oracle-says-no-critical-infrastructure-damaged-at-project-jupiter-site

Teva, Samsung Bioepis in global license, development, commercialization pact for up to 6 biosimilars

 

Teva, Samsung Bioepis sign global license, development and commercialization pact for up to six biosimilars, including Fasenra and Ilaris

• Agreement grants Teva rights to commercialize the biosimilar candidates in the U.S., Europe and Canada.
• Teva also receives options for additional commercialization territories beyond the U.S., Europe and Canada.

https://finviz.com/stock?t=TEVA&p=d

Sanofi, Regeneron expand immunology drug partnership with $1 b upfront, up to $7 b in milestones

 

• Agreement covers development of four new antibody medicines in immunology
• Deal includes settlement of a prior legal dispute between Sanofi and Regeneron

https://finviz.com/stock?t=SNY&p=d

Accenture stock surges as record bookings dispel AI fears

 Accenture (ACN) shares surged more than 20% in early trading Thursday.

Accenture's quarterly results, released on Thursday morning, eased a major concern among investors this year: that AI would squash demand for traditional consulting and IT.

The consulting firm's fiscal fourth quarter earnings per share grew 9% on an annualized basis while revenue grew 6% to $18.7 billion. Operating margin grew 370 basis points to 15.3%.

The company reached a new record high of 141 client bookings of $100 million or more in the fourth quarter, with full-year bookings reaching an all-time high of $84.5 billion.

Top-line and bottom-line growth are particularly important because they indicate strong demand for consulting services. Investors have feared AI could disrupt the industry.

Accenture has leaned into providing consulting services for AI safety, auditing, and governance.

In September, Accenture announced a partnership with Anthropic (ANT.PVT) to put dedicated safety experts directly inside the AI developer's labs to stress-test their artificial intelligence models and catch security risks or errors before they reach customers.

To fund the deal, both companies plan to invest at least $1 billion each over the next five years.

Accenture stock has been hammered this year amid worries that generative AI could automate consulting and managed services work.

The stock was down roughly 30% year to date as of Wednesday's close.

https://finance.yahoo.com/markets/article/accenture-stock-surges-as-record-bookings-dispel-ai-fears-133857510.html

French warship escorts 10 ships through Bab el-Mandeb, army spokesperson says

 

A French warship taking part in the European Union’s Red Sea mission escorted about 10 ships through the Bab el-Mandeb Strait over the past week, French army spokesperson Xavier Rival said on Thursday.

France is still discussing with Saudi partners the requirements for a possible military mission to protect energy infrastructure, he said.

https://www.iranintl.com/en/202610014523

'Iran disappointed after being excluded from Syria-Hezbollah talks in Turkey – Reuters'

 

Iran was disappointed it was not directly represented at a meeting between Syrian officials and Hezbollah in Turkey last month, Reuters reported, citing a Lebanese security source, as Ankara sought to ease tensions between the two sides.

The Lebanese security source said Hezbollah attended with Tehran’s knowledge, but Iran was disappointed not to participate directly and that Syrian President Ahmed al-Sharaa had rejected the prospect of direct ties with Iran for now. This was the main reason no agreement was reached, the source said.

The talks were the first known face-to-face meeting between the Syrian government and the Iran-backed group. They were held secretly at the invitation of Turkish security and intelligence agencies, Reuters reported, citing six sources briefed on the discussions.

The sides discussed security concerns, including Damascus’s accusations that Hezbollah was transferring weapons through Syria and its request that the group help stop smuggling and dismantle remaining cells in southern Syria. A Syrian official said Damascus assured Hezbollah it would not intervene militarily in Lebanon to disarm it. Hezbollah, in turn, pledged not to interfere in Syria’s affairs, the official said.

No final agreement was reached, and no public announcement was planned, the sources told Reuters.

Tensions between Damascus and Hezbollah rose after Bashar al-Assad was ousted in late 2024. Hezbollah had sent fighters to Syria during the civil war to support Assad’s government. Tehran was absent from the meeting, despite being Hezbollah’s main sponsor.

https://www.iranintl.com/en/202610017932