Connect with us

Published

on

Sundar Pichai, chief executive officer of Alphabet Inc., during the Google I/O Developers Conference in Mountain View, California, on Wednesday, May 10, 2023.

David Paul Morris | Bloomberg | Getty Images

Google’s new large language model, which the company announced last week, uses almost five times as much training data as its predecessor from 2022, allowing its to perform more advanced coding, math and creative writing tasks, CNBC has learned.

PaLM 2, the company’s new general-use large language model (LLM) that was unveiled at Google I/O, is trained on 3.6 trillion tokens, according to internal documentation viewed by CNBC. Tokens, which are strings of words, are an important building block for training LLMs, because they teach the model to predict the next word that will appear in a sequence.

Google’s previous version of PaLM, which stands for Pathways Language Model, was released in 2022 and trained on 780 billion tokens.

While Google has been eager to showcase the power of its artificial intelligence technology and how it can be embedded into search, emails, word processing and spreadsheets, the company has been unwilling to publish the size or other details of its training data. OpenAI, the Microsoft-backed creator of ChatGPT, has also kept secret the specifics of its latest LLM called GPT-4.

The reason for the lack of disclosure, the companies say, is the competitive nature of the business. Google and OpenAI are rushing to attract users who may want to search for information using conversational chatbots rather than traditional search engines.

But as the AI arms race heats up, the research community is demanding greater transparency.

Since unveiling PaLM 2, Google has said the new model is smaller than prior LLMs, which is significant because it means the company’s technology is becoming more efficient while accomplishing more sophisticated tasks. PaLM 2, according to internal documents, is trained on 340 billion parameters, an indication of the complexity of the model. The initial PaLM was trained on 540 billion parameters.

Google didn’t immediately provide a comment for this story.

A.I. takes center stage at Alphabet's annual Google I/O conference

Google said in a blog post about PaLM 2 that the model uses a “new technique” called “compute-optimal scaling.” That makes the the LLM “more efficient with overall better performance, including faster inference, fewer parameters to serve, and a lower serving cost.”

In announcing PaLM 2, Google confirmed CNBC’s previous reporting that the model is trained on 100 languages and performs a broad range of tasks. It’s already being used to power 25 features and products, including the company’s experimental chatbot Bard. It’s available in four sizes, from smallest to largest: Gecko, Otter, Bison and Unicorn. 

PaLM 2 is more powerful than any existing model, based on public disclosures. Facebook’s LLM called LLaMA, which it announced in February, is trained on 1.4 trillion tokens. The last time OpenAI shared ChatGPT’s training size was with GPT-3, when the company said it was trained on 300 billion tokens at the time. OpenAI released GPT-4 in March, and said it exhibits “human-level performance” on many professional tests.

LaMDA, a conversation LLM that Google introduced two years ago and touted in February alongside Bard, was trained on 1.5 trillion tokens, according to the latest documents viewed by CNBC.

As new AI applications quickly hit the mainstream, controversies surrounding the underlying technology are getting more spirited.

El Mahdi El Mhamdi, a senior Google Research scientist, resigned in February over the company’s lack of transparency. On Tuesday, OpenAI CEO Sam Altman testified at a hearing of the Senate Judiciary subcommittee on privacy and technology, and agreed with lawmakers that a new system to deal with AI is needed.

“For a very new technology we need a new framework,” Altman said. “Certainly companies like ours bear a lot of responsibility for the tools that we put out in the world.”

— CNBC’s Jordan Novet contributed to this report.

WATCH: OpenAI CEO Sam Altman calls for A.I. oversight

OpenAI CEO Sam Altman call fors A.I. oversight in testimony to congress

Continue Reading

Technology

Oracle stock jumps after $30 billion annual cloud deal revealed in filing

Published

on

By

Oracle stock jumps after  billion annual cloud deal revealed in filing

Oracle CEO Safra Catz speaks at the FII PRIORITY Summit in Miami Beach, Florida, on Feb. 20, 2025.

Joe Raedle | Getty Images

Oracle shares jumped more than 5% after a recent filing showed a cloud deal that would add over $30 billion annually.

CEO Safra Catz is slated to share the deal news at a company meeting Monday, according to a filing with the Securities and Exchange Commission. The revenues are expected to start hitting in the 2028 fiscal year.

“Oracle is off to a strong start in FY26,” Catz is expected to say, according to the filing. “Our MultiCloud database revenue continues to grow at over 100%, and we signed multiple large cloud services agreements including one that is expected to contribute more than $30 billion in annual revenue starting in FY28.”

The deals revealed Monday by Catz will not affect the company’s 2026 guidance, according to the filing.

Read more CNBC tech news

Oracle shares hit record high

Continue Reading

Technology

Trump says he has group of ‘very wealthy people’ ready to buy TikTok

Published

on

By

Trump says he has group of ‘very wealthy people’ ready to buy TikTok

U.S. President Donald Trump announced on April 4 that he would again postpone enforcement of a law banning TikTok unless its Chinese owner ByteDance divests from the platform.

Vcg | Visual China Group | Getty Images

U.S. President Donald Trump told Fox News in an interview aired on Sunday that he has a group of “very wealthy people” ready to buy TikTok, whose identities he can reveal in about two weeks.

Trump added that the deal will probably need Beijing’s approval to move forward, but said “I think President Xi will probably do it,” in reference to China’s leader Xi Jinping.

The president made the off-the-cuff remarks while discussing the possibility of another pause of his “reciprocal” tariffs on Fox News’ “Sunday Morning Futures with Maria Bartiromo.” 

Tiktok’s fate in the U.S. has been in doubt since the approval of a law in 2024 that sought to ban the platform unless its Chinese owner, ByteDance, divested from it. The legislation was driven by concerns that the Chinese government could manipulate content and access sensitive data from American users.

Earlier this month, Trump extended the deadline for ByteDance to divest from the platform’s U.S. business. It was his third extension since the Supreme Court upheld the TikTok law just a few days before Trump’s second presidential inauguration in January. The new deadline is Sept. 17. 

The Protecting Americans from Foreign Adversary Controlled Applications Act, of PAFACA, had originally been set to take effect on Jan. 19, after which app store operators and internet service providers would be penalized for supporting TikTok.

TikTok went dark in the U.S. ahead of the original deadline, but was restored after Trump provided it with assurances on the extension.

Trump, who credited the app with boosting his support among young voters in the last presidential election, has maintained that he would like to see the platform stay afloat under new ownership. 

Potential buyers that have voiced interest in the app include Trump insiders such as Oracle’s Larry Ellison to firms like AppLovin and Perplexity AI

Most of the potential bidders for TikTok don't fit both Washington and Beijing's requirements

However, it’s unclear if ByteDance would be willing to sell the company. Any potential divestiture is likely to require approval from the Chinese government.

A deal that would have spun off TikTok’s U.S. operations and allowed ByteDance to retain a minority position had been in the works in April, but was derailed by the announcement of Donald Trump’s tariffs on China, Reuters reported that month.

The president previously floated a proposal for American stakeholders to buy the company and then sell a 50% stake to the U.S. government as part of a joint venture

Experts have previously told CNBC that any potential deal could face legal challenges in the U.S., depending on whether it complies with PAFACA.

Continue Reading

Technology

Nvidia insiders dump more than $1 billion in stock, according to report

Published

on

By

Nvidia insiders dump more than  billion in stock, according to report

NVIDIA founder and CEO Jensen Huang speaks during the NVIDIA GTC Paris keynote, part of the 9th edition of the VivaTech technology startup and innovation fair, held at the Dôme de Paris in the Porte de Versailles exhibition center in Paris on June 11, 2025.

Mustafa Yalcin | Anadolu | Getty Images

Insiders at artificial intelligence chipmaker Nvidia have dumped more than $1 billion in stock over the last year, according to a report from the Financial Times.

About $500 million worth of sales occurred over the last month as the market notched new highs and shook off geopolitical tensions that had rattled investors, according to the report. The stock is up more than 17% this year despite concerns over curbs limiting AI chip sales overseas and 44% over the last three months.

Securities filings revealed that the tech titan recently unloaded about $15 million worth of shares as part of his more than $900 million plan announced in March to sell up to 6 million shares through the end of the year. Huang’s net worth totals about $138 billion, placing him as 11th on the Bloomberg Billionaires Index.

Last week, the chipmaking giant hit a fresh record and rallied for five straight days following the stock sales and an annual shareholder meeting, where the CEO called robotics the biggest opportunity for the company after AI. That helped the chipmaker regain its seat as the most valuable company ahead Microsoft and Apple.

The FT article cited a report from VerityData, which noted that the jump in shares above $150 prompted the stock dump.

Last year, Huang unloaded more than $700 million in Nvidia shares as part of a prearranged plan.

A Nvidia spokesperson declined to comment on the report.

Read the complete Financial Times report here.

Continue Reading

Trending