AI startup Anthropic has demonstrated a new technique to prevent users from eliciting harmful content from its models, as leading tech groups including Microsoft and Meta race to find ways that protect against dangers posed by the cutting-edge technology. From a report: In a paper released on Monday, the San Francisco-based startup outlined a new system called "constitutional classifiers." It is a model that acts as a protective layer on top of large language models such as the one that powers Anthropic's Claude chatbot, which can monitor both inputs and outputs for harmful content.

The development by Anthropic, which is in talks to raise $2 billion at a $60 billion valuation, comes amid growing industry concern over "jailbreaking" -- attempts to manipulate AI models into generating illegal or dangerous information, such as producing instructions to build chemical weapons. Other companies are also racing to deploy measures to protect against the practice, in moves that could help them avoid regulatory scrutiny while convincing businesses to adopt AI models safely. Microsoft introduced "prompt shields" last March, while Meta introduced a prompt guard model in July last year, which researchers swiftly found ways to bypass but have since been fixed.

An anonymous reader quotes a report from TechCrunch: TechCrunch gathered data from several sources and found similar trends. In 2024, 966 startups shut down, compared to 769 in 2023, according to Carta. That's a 25.6% increase. One note on methodology: Those numbers are for U.S.-based companies that were Carta customers and left Carta due to bankruptcy or dissolution. There are likely other shutdowns that wouldn't be accounted for through Carta, estimates Peter Walker, Carta's head of insights. [...] Meanwhile, AngelList found that 2024 saw 364 startup winddowns, compared to 233 in 2023. That's a 56.2% jump. However, AngelList CEO Avlok Kohli has a fairly optimistic take, noting that winddowns "are still very low relative to the number of companies that were funded across both years."

Layoffs.fyi found a contradicting trend: 85 tech companies shut down in 2024, compared to 109 in 2023 and 58 in 2022. But as founder Roger Lee acknowledges, that data only includes publicly reported shutdowns "and therefore represents an underestimate." Of those 2024 tech shutdowns, 81% were startups, while the rest were either public companies or previously acquired companies that were later shut down by their parent organizations. So many companies got funded in 2020 and 2021 at heated valuations with famously thin diligence, that it's only logical that up to three years later, an increasing number couldn't raise more cash to fund their operations. Taking investment at too high of a valuation increases the risk such that investors won't want to invest more unless business is growing extremely well. [...]

Looking ahead, Walker also expects we'll continue to see more shutdowns in the first half of 2025, and then a gradual decline for the rest of the year. That projection is based mostly on a time-lag estimate from the peak of funding, which he estimates was the first quarter of 2022 in most stages. So by the first quarter of 2025, "most companies will have either found a new path forward or had to make this difficult choice." "Tech zombies and a startup graveyard will continue to make headlines," said Dori Yona, CEO and co-founder of SimpleClosure. "Despite the crop of new investments, there are a lot of companies that have raised at high valuations and without enough revenue."

2021: "Automotive Startup Canoo Debuts a Snub-Nosed Electric Pickup"
2025: Canoo "Goes Belly-Up After Moving to Texas"

"Its production volumes paled in comparison to Canoo's rate of cash burn, which was substantial, with net losses in 2023 totaling just over $300 million..." reports AutoWeek. "It was able to deliver small batches of vans to a few customers, but apparently remained distant from anything approaching volume production."

"Back in 2020, electric vehicle maker Canoo snagged a $2.4 billion valuation before it had shipped a single car," remembers SFGate. "Now, just months after yanking its headquarters from Los Angeles County to Texas, the company has gone belly-up." In its four-year span as a public company, Canoo battled investor lawsuits, Securities and Exchange Commission charges, executive departures and a mixed reception of its cars. Auto tech blogger Steven Symes recently likened Canoo's cargo-style van to an "eraser on wheels."
"Canoo is the latest EV startup to go bankrupt after merging with a special purpose acquisition company (SPAC) as a shortcut to going public," notes TechCrunch. "Electric Last Mile Solutions was the first in June 2022. But since then, Fisker, Lordstown Motors, Proterra, Lion Electric, and Arrival all filed for different levels of bankruptcy protection in their various home countries." In the years since it went public, [Canoo] made a small number of its bubbly electric vans and handed them over to partners — some paying — willing to trial the vehicles. The U.S. Postal Service, Department of Defense, and NASA all have or had Canoo vehicles.

EFishery, one of Indonesia's most prominent startups, may have inflated its revenue and profit over several years, according to an internal investigation triggered by a whistleblower's claim about the company's accounting. Bloomberg News: A preliminary, ongoing probe into the agritech startup, backed by investors including SoftBank and Temasek, estimates that management inflated revenue by almost $600 million in the nine months through September last year, according to a 52-page draft report circulated among investors and reviewed by Bloomberg News. That would mean more than 75% of the reported figures were fake, the report said.

EFishery, which deploys feeders to fish and shrimp farmers in Indonesia, was a darling of the nation's startup scene and scored a valuation of $1.4 billion when G42, an AI firm controlled by United Arab Emirates royal Sheikh Tahnoon bin Zayed Al Nahyan, backed its latest funding round. It has raised hundreds of millions of dollars in an attempt to modernize the country's fish industry, providing farmers with smart feeding devices as well as feed, and then buying their produce to sell into the broader market.

An anonymous reader quotes a report from VentureBeat: Colossal BioSciences has raised $200 million in a new round of funding to bring back extinct species like the woolly mammoth. Dallas- and Boston-based Colossal is making strides in the scientific breakthroughs toward "de-extinction," or bringing back extinct species like the woolly mammoth, thylacine and the dodo. [...] Since launching in September 2021, Colossal has raised $435 million in total funding. This latest round of capital places the company at a $10.2 billion valuation. Colossal will leverage this latest infusion of capital to continue to advance its genetic engineering technologies while pioneering new revolutionary software, wetware and hardware solutions, which have applications beyond de-extinction including species preservation and human healthcare.

"Our recent successes in creating the technologies necessary for our end-to-end de-extinction toolkit have been met with enthusiasm by the investor community. TWG Global and our other partners have been bullish in their desire to help us scale as quickly and efficiently as possible," said CEO Colossal Ben Lamm, in a statement. "This funding will grow our team, support new technology development, expand our de-extinction species list, while continuing to allow us to carry forth our mission to make extinction a thing of the past." Here's a summary of the startup's progress on its efforts to bring back the woolly mammoth, thylacine and the dodo:

Woolly Mammoth De-extinction Progress
- Generated chromosome-scale reference genomes for elephants and the first de novo assembled mammoth genome
- Acquired and aligned 60+ ancient mammoth genomes and 30+ genomes of extant elephant species, improving mammoth-specific variant accuracy
- Derived pluripotent stem cells for Asian elephants, advancing reproductive technologies essential for de-extinction

Thylacine De-extinction Progress
- Created a 99.9% complete ancient genome for the thylacine using long-read and RNA sequencing
- Assembled telomere-to-telomere genomes of dasyurid species to understand evolutionary relationships and support conservation of marsupials
- Progress in genomics and reproductive technologies positions Colossal ahead of schedule on critical de-extinction steps

Dodo De-extinction Progress
- Completed high-coverage genomes for the dodo, its relatives, and the critically endangered manumea
- Developed tools for avian genome engineering, including techniques for craniofacial gene-editing and primordial germ cell cultivation
- Significant advances in avian-specific genetic techniques are driving progress toward dodo restoration and bird conservation

An anonymous reader shares a report: Chinese venture capitalists are hounding failed founders [non-paywalled source], pursuing personal assets and adding the individuals to a national debtor blacklist when they fail to pay up, in moves that are throwing the country's startup funding ecosystem into crisis. The hard-nosed tactics by risk capital providers have been facilitated by clauses known as redemption rights, included in nearly all the financing deals struck during China's boom times.

"My investors verbally promised they wouldn't enforce them, that they had never enforced them before -- and in '17 and '18 that was true -- no one was enforcing them," said Neuroo Education founder Wang Ronghui, who now owes investors millions of dollars after her childcare chain stumbled during the pandemic.

While they are relatively rare in US venture investing, more than 80% of venture and private equity deals in China contain redemption provisions, according to Shanghai-based law firm Lifeng Partners estimates. They typically require companies, and often their founders as well, to buy back investors' shares plus interest if certain targets such as an initial public offering timeline, valuation goals or revenue metrics are not met.

Anthropic is in advanced talks to raise $2 billion dollars in a deal that would value it at $60 billion, making it the latest artificial-intelligence startup to seize upon investor euphoria for the technology. WSJ: The funding round is being led by the venture firm Lightspeed Venture Partners, people familiar with the matter said. The $60 billion valuation includes the money Anthropic plans to raise in the round.

The deal would make Anthropic the fifth-most valuable U.S. startup after SpaceX, OpenAI, Stripe and Databricks, according to data provider CB Insights. It was valued last year at $18 billion in a round led by Menlo Ventures. There has been a dealmaking frenzy among AI companies since OpenAI raised $6.6 billion in an October round that nearly doubled its value to $157 billion. Two other startups, Elon Musk's xAI and Perplexity, subsequently raised money at substantially increased valuations.

Slashdot reader DevNull127 writes: This year the London Stock Exchange got a new listing for "Raspberry Pi Holdings plc." It's the computer-making commercial subsidiary of their larger educational charity, the Raspberry Pi Foundation. "Access to the public market will enable us to build more of the products you love, faster," explained CEO Eben Upton in June. And in May Foundation head Philip Colligan added that beyond the $50 million already donated to their educational charity by the commercial subsidiary, the IPO would allow the conversion of some stock sales to "an endowment that we will use to fund our educational programmes... The Foundation will use any funds that we raise through the sale of shares at the IPO — or subsequently — to advance our ambitious global strategy to enable every young person to realise their full potential through the power of computing and digital technologies."

So how's that working out? A finance site called Proactive Investors UK reports that in September Raspbery Pi Holdings plc "reported underlying profits (EBITDA) of US$20.9 million, up by 55% from a year ago, on revenues up 61% to US$144 million... The Pi 5 single-board computer (SBC), launched at the end of last October [2023], sold 1.1 million units in the first half, with overall unit growth at 31%."

And then in December its stock price suddenly shot up to more than double where it was at the end of November — giving Raspbery Pi Holdings plc a valuation "just under £1.3 billion."

OpenAI plans to overhaul its corporate structure by converting its for-profit business into a Delaware public benefit corporation, seeking to raise capital from investors who want conventional equity stakes.

The Microsoft-backed AI startup will scrap its unusual hybrid model where a nonprofit controls a capped-profit entity. The restructuring aims to help OpenAI compete with tech giants pouring hundreds of billions into AI development, it said.

Under the plan, OpenAI's nonprofit wing will receive shares in the new public benefit corporation at a valuation set by outside advisers. The nonprofit will pursue charitable work in healthcare and education while the corporation runs OpenAI's main operations.

The startup, which launched ChatGPT in 2022 and claims 300 million weekly users, said its current structure hampers fundraising at the scale needed to advance artificial general intelligence development. The restructured business will maintain OpenAI's mission of ensuring AI benefits humanity as its legal mandate.

An anonymous reader quotes a report from Gizmodo: Once an icon of the 20th century seen as obsolete in the 21st, Encyclopedia Britannica -- now known as just Britannica -- is all in on artificial intelligence, and may soon go public at a valuation of nearly $1 billion, according to the New York Times.

Until 2012 when printing ended, the company's books served as the oldest continuously published, English-language encyclopedias in the world, essentially collecting all the world's knowledge in one place before Google or Wikipedia were a thing. That has helped Britannica pivot into the AI age, where models benefit from access to high-quality, vetted information. More general-purpose models like ChatGPT suffer from hallucinations because they have hoovered up the entire internet, including all the junk and misinformation.

While it still offers an online edition of its encyclopedia, as well as the Merriam-Webster dictionary, Britannica's biggest business today is selling online education software to schools and libraries, the software it hopes to supercharge with AI. That could mean using AI to customize learning plans for individual students. The idea is that students will enjoy learning more when software can help them understand the gaps in their understanding of a topic and stay on it longer. Another education tech company, Brainly, recently announced that answers from its chatbot will link to the exact learning materials (i.e. textbooks) they reference.

Britannica's CEO Jorge Cauz also told the Times about the company's Britannica AI chatbot, which allows users to ask questions about its vast database of encyclopedic knowledge that it collected over two centuries from vetted academics and editors. The company similarly offers chatbot software for customer service use cases. Britannica told the Times it is expecting revenue to double from two years ago, to $100 million.

"From the moment GPT-4 came out in March 2023, OpenAI has been working on GPT-5..." reports the Wall Street Journal. [Alternate URL here.] But "OpenAI's new artificial-intelligence project is behind schedule and running up huge bills. It isn't clear when — or if — it'll work."

"There may not be enough data in the world to make it smart enough." OpenAI's closest partner and largest investor, Microsoft, had expected to see the new model around mid-2024, say people with knowledge of the matter. OpenAI has conducted at least two large training runs, each of which entails months of crunching huge amounts of data, with the goal of making Orion smarter. Each time, new problems arose and the software fell short of the results researchers were hoping for, people close to the project say... [And each one costs around half a billion dollars in computing costs.]

The $157 billion valuation investors gave OpenAI in October is premised in large part on [CEO Sam] Altman's prediction that GPT-5 will represent a "significant leap forward" in all kinds of subjects and tasks.... It's up to company executives to decide whether the model is smart enough to be called GPT-5 based in large part on gut feelings or, as many technologists say, "vibes."

So far, the vibes are off...
OpenAI wants to use its new model to generate high-quality synthetic data for training, according to the article. But OpenAI's researchers also "concluded they needed more diverse, high-quality data," according to the article, since "The public internet didn't have enough, they felt." OpenAI's solution was to create data from scratch. It is hiring people to write fresh software code or solve math problems for Orion to learn from. [And also theoretical physics experts] The workers, some of whom are software engineers and mathematicians, also share explanations for their work with Orion... Having people explain their thinking deepens the value of the newly created data. It's more language for the LLM to absorb; it's also a map for how the model might solve similar problems in the future... The process is painfully slow. GPT-4 was trained on an estimated 13 trillion tokens. A thousand people writing 5,000 words a day would take months to produce a billion tokens.

OpenAI's already-difficult task has been complicated by internal turmoil and near-constant attempts by rivals to poach its top researchers, sometimes by offering them millions of dollars... More than two dozen key executives, researchers and longtime employees have left OpenAI this year, including co-founder and Chief Scientist Ilya Sutskever and Chief Technology Officer Mira Murati. This past Thursday, Alec Radford, a widely admired researcher who served as lead author on several of OpenAI's scientific papers, announced his departure after about eight years at the company...

OpenAI isn't the only company worrying that progress has hit a wall. Across the industry, a debate is raging over whether improvement in AIs is starting to plateau. Sutskever, who recently co-founded a new AI firm called Safe Superintelligence or SSI, declared at a recent AI conference that the age of maximum data is over. "Data is not growing because we have but one internet," he told a crowd of researchers, policy experts and scientists. "You can even go as far as to say that data is the fossil fuel of AI."

And that fuel was starting to run out.

They provide training data to top AI companies including OpenAI and Meta, according to its web site. Founded in 2016, San Francisco-based Scale AI now has over 900 employees, eventually growing beyond "unicorn" status with over $1.35 billion in ivnestments. In May the company's valuation was over $14 billion, with investors including Amazon, Meta, Nvidia, Cisco, Intel, and AMD (as well as earlier investments from Y Combinator and $100 million from Peter Thiel's Founders Fund). SFGate calls them "a buzzy San Francisco startup with high-dollar ties across the tech industry".

But SFGate also report Scale AI "was sued Tuesday by a former worker with allegations that the company is committing wage theft and misclassifying workers." Steve McKinney filed the suit against Scale and several top executives, including 27-year-old billionaire CEO Alexandr Wang, in San Francisco Superior Court. With the filing, the former contractor aims to be a lead plaintiff for a class-action lawsuit against Scale; a judge will need to certify his proposed class of current and former contractors within California...

McKinney, whose complaint says he was paid on an hourly basis and worked on a project eventually sold to Meta, is accusing Scale of amassing its clout and cash by exploiting workers. "Scale AI is the sordid underbelly propping up the generative AI industry," the complaint says, before rattling off a list of allegations about its treatment of contractors. The document accuses Scale of bait-and-switch hiring promises; demanding off-the-clock, unpaid work; denying overtime pay; and unfairly booting contractors from projects...

The Tuesday complaint calls Scale's control over its contractors "Orwellian." The company makes contractors download a tool to track much of their computer use, including by taking periodic screenshots, the suit alleges. The lawsuit also alleges that Scale reassigns the workers to varyingly paid projects and docks pay if a task takes more than it was supposed to, plus posits that Scale is in violation of California's "ABC" test, which monitors use of the designation "independent contractor." It argues that contracted "Taskers" like McKinney should be classified as employees instead...

The complaint, along with arguing for class-action certification, seeks restitution, punitive damages and changes to Scale's worker classification model.
The article adds that "Per Fortune, Scale's armies of contractors marked up images for Cruise and Waymo to help autonomous cars understand their surroundings..."

EvenUp, a legal tech startup valued at over $1 billion, has largely relied on human workers rather than AI to process personal injury claims, according to Business Insider, which spoke to multiple employees.

The startup, which promised to automate the analysis of medical records and case files to determine accident compensation, saw its valuation surge from $85 million to $1 billion in October. Former staff members reported that supervisors instructed them not to use the AI system due to its unreliability, with some working until 3 a.m. to complete tasks manually.

The ex-employees cited numerous AI errors, including missed injuries, fabricated medical conditions, and incorrectly recorded doctor visits. While no errors made it into final products, such mistakes could have reduced victim payouts if not caught by human reviewers, they cautioned.

An anonymous reader quotes a report from the New York Times: The founder of an artificial intelligence start-up focused on education was arrested and charged with defrauding her investors, lying about the company's profits and falsely claiming that some of the largest school districts in the country, including New York City's, were her customers. The founder, Joanna Smith-Griffin, started the company, AllHere Education, in 2016, with the goal of using artificial intelligence to increase student and parent engagement and curb absenteeism. In the years that followed, Ms. Smith-Griffin, 33, misrepresented AllHere's revenue and customer base to fraudulently raise almost $10 million in funds, according to the indictment. Once the company's valuation had climbed, she sold some of her stake in it and spent hundreds of thousands of dollars on a down payment for a new home and on her wedding.

Ms. Smith-Griffin was arrested Tuesday in North Carolina, where she lives, and charged with wire fraud, securities fraud and aggravated identity theft. She faces more than 40 years in prison. AllHere is now in bankruptcy proceedings, prosectors said, and all of its employees have been laid off. "Her alleged actions impacted the potential for improved learning environments across major school districts by selfishly prioritizing personal expenses," said James E. Dennehy, the F.B.I. assistant director in New York leading the investigation into Ms. Smith-Griffin. "The F.B.I. will ensure that any individual exploiting the promise of educational opportunities for our city's children will be taught a lesson." Smith-Griffin is the latest Forbes 30 Under 30 honoree to be indicted on fraud. "The Forbes-to-Fraud pipeline includes FTX founder Sam Bankman-Fried and Caroline Ellison, co-CEO of Alameda Research; fintech Frank founder Charlie Javice; and 'Pharma bro' Martin Shkreli," notes TechCrunch.

An anonymous reader quotes a report from Ars Technica: Anthropic has announced a partnership with Palantir and Amazon Web Services to bring its Claude AI models to unspecified US intelligence and defense agencies. Claude, a family of AI language models similar to those that power ChatGPT, will work within Palantir's platform using AWS hosting to process and analyze data. But some critics have called out the deal as contradictory to Anthropic's widely-publicized "AI safety" aims. On X, former Google co-head of AI ethics Timnit Gebru wrote of Anthropic's new deal with Palantir, "Look at how they care so much about 'existential risks to humanity.'"

The partnership makes Claude available within Palantir's Impact Level 6 environment (IL6), a defense-accredited system that handles data critical to national security up to the "secret" classification level. This move follows a broader trend of AI companies seeking defense contracts, with Meta offering its Llama models to defense partners and OpenAI pursuing closer ties with the Defense Department. In a press release, the companies outlined three main tasks for Claude in defense and intelligence settings: performing operations on large volumes of complex data at high speeds, identifying patterns and trends within that data, and streamlining document review and preparation.

While the partnership announcement suggests broad potential for AI-powered intelligence analysis, it states that human officials will retain their decision-making authority in these operations. As a reference point for the technology's capabilities, Palantir reported that one (unnamed) American insurance company used 78 AI agents powered by their platform and Claude to reduce an underwriting process from two weeks to three hours. The new collaboration builds on Anthropic's earlier integration of Claude into AWS GovCloud, a service built for government cloud computing. Anthropic, which recently began operations in Europe, has been seeking funding at a valuation up to $40 billion. The company has raised $7.6 billion, with Amazon as its primary investor.

"Nvidia dethroned Apple as the world's most valuable company on Friday..." reports Reuters, "powered by insatiable demand for its specialized artificial intelligence chips." Nvidia's stock market value briefly touched $3.53 trillion, slightly above Apple's $3.52 trillion, LSEG data showed... In June, Nvidia briefly became the world's most valuable company before it was overtaken by Microsoft and Apple. The tech trio's market capitalizations have been neck-and-neck for several months. [Friday] Microsoft's market value stood at $3.18 trillion, with its stock up 0.8%...

Nvidia's shares hit a record high on Tuesday, building on a rally from last week when TSMC, the world's largest contract chipmaker, posted a forecast-beating 54% jump in quarterly profit driven by soaring demand for chips used in AI.
The article points out that by the end of the day Friday, Nvidia's valuation had dropped to $3.47 trillion, while Apple had risen to $3.52 trillion...

Intel considered acquiring graphics chip maker Nvidia for up to $20 billion in 2005, a move that could have reshaped the AI industry, according to The New York Times. Then-CEO Paul Otellini pitched the acquisition to Intel's board, recognizing the potential of graphics processors for data center computing. The board rejected the proposal, citing Intel's poor track record with acquisitions and the deal's unprecedented size, the report added. Today, Nvidia dominates the AI chip market with a $3 trillion valuation, while Intel struggles with declining revenue and recent layoffs of 16,000 workers.

An anonymous reader shares a report: Byju Raveendran, the founder of the embattled edtech group Byju's, acknowledged on Thursday afternoon that he made mistakes, mistimed the market, overestimated growth potential and that his startup, once valued at $22 billion, is now effectively worth "zero."

Speaking to a group of journalists, Raveendran said the company's aggressive acquisition of more than two dozen startups to expand into new markets proved fatal when financing dried up in 2022. Byju's was planning to go public in early 2022 with several investment bankers giving the firm valuation as high as $50 billion, TechCrunch reported earlier.

He alleged that many of his more than 100 investors had urged him to pursue aggressive expansion into as many as 40 markets. But, he added, those very investors got cold feet when global markets tumbled following Russia's invasion of Ukraine, sending the venture capital market into a downward spiral.

An anonymous reader quotes a report from TechCrunch: Stripe is in talks to acquire stablecoin platform Bridge for a whopping $1 billion, according to Forbes (paypalled). The talks are reportedly in advanced stages, although nothing has been finalized. Bridge, co-founded by Coinbase alumni Zach Abrams and Sean Yu, has built an API that helps companies accept stablecoins. The pair raised $58 million from investors like Index Ventures and Sequoia Capital, according to PitchBook. If the deal with Stripe goes through, it would be a huge jump from Bridge's $200 million valuation, as well as being Stripe's largest acquisition to date.

Cybersecurity firm IronNet, founded by former NSA director Keith Alexander, has collapsed after failing to deliver on its promise to revolutionize cyber defense. The company, which went public in 2021 with a $3 billion valuation, shut down in September 2023 after running out of money.

IronNet's downfall has left investors and former employees bitter, with some accusing the company of misleading them about its financial health. "I'm honestly ashamed that I was ever an executive at that company," said Mark Berly, a former IronNet vice president. He said the company's top leaders cultivated a culture of deceit "just like Theranos." Critics point to questionable business practices, subpar products, and associations that potentially exposed the firm to Russian influence. The company's board included high-profile national security figures, which helped attract investments and contracts. However, IronNet struggled to secure major deals and meet revenue projections.