Tag:

Model

Coinbase to Open New San Francisco Office After Dropping HQ Model

by admin May 30, 2025

Crypto exchange Coinbase has signed a lease for office space in San Francisco’s Mission Rock development, marking the crypto exchange’s return to the city after abandoning its headquarters model roughly three years ago.

The move is based on a deal Coinbase secured for a 150,000 square foot spot at 1090 Dr. Maya Angelou Lane, and represents more than half of Building B at the waterfront development built by Tishman Speyer and the San Francisco Giants.

San Francisco “is the place to build and grow,” Mayor Daniel Lurie wrote Thursday afternoon on X, welcoming Coinbase.

The space will serve as Coinbase’s largest single office, according to a report by The San Francisco Standard, cited by Mayor Lurie.

Responding to Mayor Lurie, Coinbase CEO Brian Armstrong said that there was “still lots of work to do to improve the city,” saying that it was “badly run for many years,” later adding that Mayor Lurie’s work “has not gone unnoticed.”

Two weeks earlier, Coinbase was listed in the S&P 500, cementing its stature among publicly listed companies in the U.S.

Tax concerns, lease obligations

Coinbase’s return to the city comes after it paid $25 million for the “early termination of an office lease,” its 2023 shareholder letter shows, following its transition to operating without a traditional headquarters in February 2021.

Coinbase announced on May 5, 2021, that it would close its San Francisco office—its former headquarters—in 2022 as part of its transition to a remote-first model. The company emphasized that this move was intended to ensure no single location would be considered its headquarters, aligning with its decentralized workforce strategy.

While the exact closure date in 2022 was not publicly specified, the office at 430 California Street was fully decommissioned that year.

In 2022, a similar move out of the city was made by Kraken, a rival U.S. exchange, with its former CEO Jesse Powell saying the city had “fallen quite far,” recounting how he has seen it “deteriorate” since he moved in 2013.

“We never left California. Lots of our employees live there. We go to where the talent is,” Armstrong said in response to concerns pointing out that San Francisco was a “tax-heavy state.”

Data compiled by the tax consulting and advisory firm Ryan indicates that businesses in San Francisco pay taxes based on their annual revenue, with two central taxes that become more expensive as companies grow their earnings.

Small businesses that earn less than $5 million annually are exempt from the main business tax, but larger companies can pay up to approximately 4% of their total revenue when both taxes are combined, according to changes in business tax for the city approved in November of last year.

Coinbase’s 10-K SEC filing for late 2024 describes the company as a “remote-first company” that does not “maintain a headquarters” for its roughly 3,800 employees.

The report indicates $132.3 million in global total operating lease obligations for corporate offices, with $9.9 million due in the next 12 months.

Edited by Sebastian Sinclair

Daily Debrief Newsletter

Start every day with the top news stories right now, plus original features, a podcast, videos and more.

Source link

May 30, 2025 0 comments

Esports

CFP 5+11 model gaining traction as leaders eye next steps

by admin May 30, 2025

DESTIN, Fla. — A 16-team College Football Playoff model featuring the top five conference champions and 11 at-large teams is gaining traction following SEC spring meetings this week, but the next step in playoff expansion for 2026 and beyond will depend on how quickly the sport’s leaders can make a flurry of decisions.

A critical component is the SEC’s choice between staying at eight league games or moving to nine, a topic ACC sources say could be revisited in their league after years of being dormant if prompted by playoff expansion. The linchpin to those scheduling decisions is one thing every conference seems to agree on: the need for clarity about how the CFP selection committee ranks its teams, starting with how strength of schedule is determined and applied.

“I do think there’s a need for change,” SEC commissioner Greg Sankey said of the ranking protocol Thursday at the conclusion of his league’s spring meetings. “… How do you make those decisions? It’s hard, and we trust the committee to do that, and I respect the people in there, so this isn’t a criticism of the people. This is wanting to understand the decisions. We have to have better clarity on the criteria that inform those decisions.”

Editor’s Picks

2 Related

Currently, strength of schedule is one of several factors not weighed in the committee’s ambiguous protocol — language the FBS commissioners wrote at the inception of the four-team playoff in 2014. There’s a sense among some athletic directors in the SEC and ACC that moving to nine conference games is feasible — if the committee doesn’t penalize teams for losing two or three games against strong opponents.

Some SEC athletic directors stressed this week that they would only favor a nine-game league schedule if the conference is guaranteed four playoff spots — also the Big Ten’s preferred model.

“If we’re not confident that the decision-making about who gets in and why and what are the metrics around it, it’s going to be really hard for some of my colleagues to get to the nine games,” Texas A&M athletic director Trev Alberts said. “We’ve got a timeline that’s getting tight, and we recognize that. It seems like everything is coming to a head. In a way it’s a little bit frustrating, in another way it feels good because eventually, it feels like we’re actually going to get some of this dealt with.”

CFP leaders have set Dec. 1 as a deadline to determine the future format, and Sankey said he wants to make a scheduling decision in 2026, but didn’t specify when. The FBS commissioners and Notre Dame athletic director Pete Bevacqua are scheduled to meet in person June 18 in Asheville, North Carolina.

Sankey was asked if his conference will be unified on a format by then.

“We’ll see,” Sankey said.

Multiple ACC sources said the conference would prefer a 5+11 model, and Big 12 commissioner Brett Yormark has publicly supported it at his league’s spring meetings this week.

“It has always been our first choice,” Yormark told ESPN. “It’s fair and rewards on-field performance. I’m not surprised SEC coaches like it.”

The Big 12’s administrators agree.

“The construct of the CFP wasn’t to give one or two conferences more value. It was supposed to be the best way to conduct a real national championship,” UCF athletic director Terry Mohajir said. “I think a 5+11 is the best way to do that, and it gets the best teams in.”

If Sankey can get his athletic directors on the same page as his coaches, who this week voiced strong support for a 5+11 model (but with eight conference games), the Big Ten would likely be the lone league in the room pushing automatic qualifiers.

“[We’re] kind of important,” Sankey said, “a bit important in that decision.”

The Big Ten and SEC have the bulk of control over the playoff’s format in 2026 and beyond, something the other FBS commissioners and Bevacqua agreed to when they signed a memorandum of understanding for the new six-year deal.

“If we do want to have a national tournament, we do have to get everyone on the same page and everyone has to work together,” Florida athletic director Scott Stricklin said.

The ACC’s considerations of a nine-game league schedule had been tabled for a few years for multiple reasons. Several schools already have existing rivalries with SEC schools, plus there is a built-in agreement with Notre Dame. The ACC doesn’t necessarily have to decide that for the 2026 season. It’s something that could be phased in, according to a source.

Alabama athletic director Greg Byrne said two of his main priorities as the playoff discussions move forward are access and having “a pretty clear understanding of what gets you in, what doesn’t.”

“I know last year I talked about a lot of what I read was two versus three losses, and that was concerning,” Byrne said. “Granted, ultimately, it’s up to you and the play you have on the field, and you have to recognize that, but I also do believe that when you looked at the bullet points for the CFP, strength of schedule was the first bullet point listed. Trying to get some clear understanding of how is that weighed in the room is important. Our conference because of the play on the field has deserved the benefit of the doubt at times to be strongly considered for the CFP.”

On Thursday afternoon, the SEC provided members of the media with a six-page packet that included color-coded charts using multiple metrics to illustrate the league’s dominant schedule strength. Sankey said the task for determining the CFP’s strength of schedule component is striking a balance “between human and machine,” referring to the old BCS computer formula.

“Whether you agree or not, that’s what we’re looking at,” Sankey said of the packet, which included ESPN’s Strength of Record, Bill Connelly’s SP+, Kenneth Massey’s metric, ESPN’s Football Power Index and ESPN’s Strength of Schedule metric. “That doesn’t mean every one of these should be inserted into the CFP, but I think you have to consider what it means, because there’s other ratings and evaluation tools we’ve looked at that are much like these results.”

While the issues are on the table, the CFP’s management committee is notorious for missing its own deadlines. Sankey this week didn’t rule out the possibility of the 12-team CFP remaining in place in 2026.

“Can I see a scenario? Sure, I can see a scenario,” Sankey said. “But is that the most likely scenario? Come back for more. I said — genuinely — we’re interested in a model. We’re not committed to that model, and you’ve seen that play out this week, where people have different ideas.”

The question is if the Power 4 leagues can put their differences aside — quickly.

“We need to work well together,” Sankey said. “The emotional maturity needed right now is higher than ever.”

ESPN’s Andrea Adelson contributed to this report.

Source link

May 30, 2025 0 comments

Crypto Trends

DeepSeek Claims Upgraded Model Approaching ChatGPT, Gemini

by admin May 29, 2025

DeepSeek, a China-based artificial intelligence company, has announced an upgrade to its AI chatbot, saying it can now offer enhanced overall logic, mathematics and programming with a reduced hallucination rate.

According to DeepSeek, the upgraded model — DeepSeek-R1-0528 — has “significantly improved its depth of reasoning and inference capabilities.” The startup said the model’s overall performance is now “approaching that of leading models, such as O3 and Gemini 2.5 Pro.”

Performance comparison of language models across six benchmarks. Source: DeepSeek

DeepSeek’s debut of its R1 chatbot in January sent shockwaves through the AI industry and further established China as an AI force. The company’s first AI model had a training cost of $6 million and similar performance to leading AI models trained on significantly larger sums of capital.

According to data from Business of Apps, DeepSeek has been downloaded 75 million times since its launch and had 38 million monthly active users (MAU) as of April. In a recent antitrust lawsuit, Google estimated that Gemini reached 350 million active users in March, while OpenAI’s ChatGPT claimed 600 million active users in the same month.

Related: China’s DeepSeek launches new open-source AI after R1 took on OpenAI

Chinese-American AI race heats up

The United States government is planning to restrict the sale of advanced chip design software to China. According to a Bloomberg report, the move seeks to limit China’s ability to advance its domestic semiconductor manufacturing capabilities.

Semiconductors are critical for a wide range of technologies, including AI, where they serve as the hardware backbone for training and running complex models.

New China AI models, such as Tencent’s T1 and Alibaba’s Qwen3, have also emerged in the first few months of 2025, spurring the AI race along.

Magazine: AI Eye: 9 curious things about DeepSeek R1

Source link

May 29, 2025 0 comments

Gaming Gear

Why Anthropic’s New AI Model Sometimes Tries to ‘Snitch’

by admin May 28, 2025

The hypothetical scenarios the researchers presented Opus 4 with that elicited the whistleblowing behavior involved many human lives at stake and absolutely unambiguous wrongdoing, Bowman says. A typical example would be Claude finding out that a chemical plant knowingly allowed a toxic leak to continue, causing severe illness for thousands of people—just to avoid a minor financial loss that quarter.

It’s strange, but it’s also exactly the kind of thought experiment that AI safety researchers love to dissect. If a model detects behavior that could harm hundreds, if not thousands, of people—should it blow the whistle?

“I don’t trust Claude to have the right context, or to use it in a nuanced enough, careful enough way, to be making the judgment calls on its own. So we are not thrilled that this is happening,” Bowman says. “This is something that emerged as part of a training and jumped out at us as one of the edge case behaviors that we’re concerned about.”

In the AI industry, this type of unexpected behavior is broadly referred to as misalignment—when a model exhibits tendencies that don’t align with human values. (There’s a famous essay that warns about what could happen if an AI were told to, say, maximize production of paperclips without being aligned with human values—it might turn the entire Earth into paperclips and kill everyone in the process.) When asked if the whistleblowing behavior was aligned or not, Bowman described it as an example of misalignment.

“It’s not something that we designed into it, and it’s not something that we wanted to see as a consequence of anything we were designing,” he explains. Anthropic’s chief science officer Jared Kaplan similarly tells WIRED that it “certainly doesn’t represent our intent.”

“This kind of work highlights that this can arise, and that we do need to look out for it and mitigate it to make sure we get Claude’s behaviors aligned with exactly what we want, even in these kinds of strange scenarios,” Kaplan adds.

There’s also the issue of figuring out why Claude would “choose” to blow the whistle when presented with illegal activity by the user. That’s largely the job of Anthropic’s interpretability team, which works to unearth what decisions a model makes in its process of spitting out answers. It’s a surprisingly difficult task—the models are underpinned by a vast, complex combination of data that can be inscrutable to humans. That’s why Bowman isn’t exactly sure why Claude “snitched.”

“These systems, we don’t have really direct control over them,” Bowman says. What Anthropic has observed so far is that, as models gain greater capabilities, they sometimes select to engage in more extreme actions. “I think here, that’s misfiring a little bit. We’re getting a little bit more of the ‘Act like a responsible person would’ without quite enough of like, ‘Wait, you’re a language model, which might not have enough context to take these actions,’” Bowman says.

But that doesn’t mean Claude is going to blow the whistle on egregious behavior in the real world. The goal of these kinds of tests is to push models to their limits and see what arises. This kind of experimental research is growing increasingly important as AI becomes a tool used by the US government, students, and massive corporations.

And it isn’t just Claude that’s capable of exhibiting this type of whistleblowing behavior, Bowman says, pointing to X users who found that OpenAI and xAI’s models operated similarly when prompted in unusual ways. (OpenAI did not respond to a request for comment in time for publication).

“Snitch Claude,” as shitposters like to call it, is simply an edge case behavior exhibited by a system pushed to its extremes. Bowman, who was taking the meeting with me from a sunny backyard patio outside San Francisco, says he hopes this kind of testing becomes industry standard. He also adds that he’s learned to word his posts about it differently next time.

“I could have done a better job of hitting the sentence boundaries to tweet, to make it more obvious that it was pulled out of a thread,” Bowman says as he looked into the distance. Still, he notes that influential researchers in the AI community shared interesting takes and questions in response to his post. “Just incidentally, this kind of more chaotic, more heavily anonymous part of Twitter was widely misunderstanding it.”

Source link

May 28, 2025 0 comments