Connect with us

Published

on

Anthropic on Monday debuted Claude 3, a suite of artificial intelligence models that it says are its fastest and most powerful yet. The new tools are called Claude 3 Opus, Sonnet and Haiku.

The company said the most capable of the new models, Claude 3 Opus, outperformed OpenAI’s GPT-4 and Google’s Gemini Ultra on industry benchmark tests, such as undergraduate level knowledge, graduate level reasoning and basic mathematics.

This is the first time Anthropic has offered multimodal support. Users can upload photos, charts, documents and other types of unstructured data for analysis and answers.

The other models, Sonnet and Haiku, are more compact and less expensive than Opus. Sonnet and Opus are available in 159 countries starting Monday, while Haiku will be coming soon, according to Anthropic. The company declined to specify how long it took to train Claude 3 or how much it cost, but it said companies like Airtable and Asana helped A/B test the models.

This time last year, Anthropic was seen as a promising generative AI startup founded by ex-OpenAI research executives. It had completed Series A and B funding rounds, but it had only rolled out the first version of its chatbot without any consumer access or major fanfare.

Twelve months later, it’s one of the hottest AI startups, with backers including Google, Salesforce and Amazon, and a product that directly competes with ChatGPT in both the enterprise and consumer worlds. Over the past year, the startup closed five different funding deals, totaling about $7.3 billion.

The generative AI field has exploded over the past year, with a record $29.1 billion invested across nearly 700 deals in 2023, a more than 260% increase in deal value from a year earlier, according to PitchBook. It’s become the buzziest phrase on corporate earnings calls quarter after quarter. Academics and ethicists have voiced significant concerns about the technology’s tendency to propagate bias, but even so, it’s quickly made its way into schools, online travel, the medical industry, online advertising and more.

Between 60 and 80 people worked on the core AI model, while between 120 and 150 people worked on its technical aspects, Anthropic co-founder Daniela Amodei told CNBC in an interview. For the AI model’s last iteration, a team of 30 to 35 people worked directly on it, with about 150 people total supporting it, Amodei told CNBC in July.

Anthropic said Claude 3 can summarize up to about 150,00 words, or a sizeable book (think: around the length range of “Moby Dick” or “Harry Potter and the Deathly Hallows”). Its previous version could only summarize 75,000 words. Users can input large data sets, and ask for summaries in the form of a memo, letter or story. ChatGPT, by contrast, can handle about 3,000 words.

Amodei also said Claude 3 has a better understanding of risk in responses than its previous version.

“In our quest to have a highly harmless model, Claude 2 would sometimes over-refuse,” Amodei told CNBC. “When somebody would kind of bump up against some of the spicier topics or the trust and safety guardrails, sometimes Claude 2 would trend a little bit conservative in responding to those questions.”

Claude 3 has a more nuanced understanding of prompts, according to Anthropic.

Multimodality, or adding options like photo and video capabilities to generative AI, whether uploading them yourself or creating them using an AI model, has quickly become one of the industry’s hottest use cases.

“The world is multimodal,” OpenAI COO Brad Lightcap told CNBC in November. “If you think about the way we as humans process the world and engage with the world, we see things, we hear things, we say things — the world is much bigger than text. So to us, it always felt incomplete for text and code to be the single modalities, the single interfaces that we could have to how powerful these models are and what they can do.”

But multimodality, and increasingly complex AI models, also lead to more potential risks. Google recently took its AI image generator, part of its Gemini chatbot, offline after users discovered historical inaccuracies and questionable responses, which have circulated widely on social media.

Anthropic’s Claude 3 does not generate images; instead, it only allows users to upload images and other documents for analysis.

“Of course no model is perfect, and I think that’s a very important thing to say upfront,” Amodei told CNBC. “We’ve tried very diligently to make these models the intersection of as capable and as safe as possible. Of course there are going to be places where the model still makes something up from time to time.”

Clarification: Anthropic clarified with CNBC that Claude 3 can summarize about 150,000 words, not 200,000.

Continue Reading

Technology

South Korea says DeepSeek transferred user data to China and the U.S. without consent

Published

on

By

South Korea says DeepSeek transferred user data to China and the U.S. without consent

Jaap Arriens | Nurphoto | Getty Images

South Korea’s data protection authority has concluded that Chinese artificial intelligence startup DeepSeek collected personal information from local users and transferred it overseas without their permission.

The authority, the Personal Information Protection Commission, released its written findings on Thursday in connection with a privacy and security review of DeepSeek.

It follows DeepSeek’s removal of its chatbot application from South Korean app stores in February at the recommendation of PICP. The agency said DeepSeek had committed to cooperate on its concerns

During DeepSeek’s presence in South Korea, it transferred user data to several firms in China and the U.S. without obtaining the necessary consent from users or disclosing the practice, the PIPC said.

The agency highlighted a particular case in which DeepSeek transferred information from user-written AI prompts, as well as device, network, and app information, to a Chinese cloud service platform named Beijing Volcano Engine Technology Co.

While the PIPC identified Beijing Volcano Engine Technology Co. as “an affiliate” of TikTok-owner ByteDance, the information privacy watchdog noted in a statement that the cloud platform “is a separate legal entity and has no relation to ByteDance,” according to a Google translation.

According to PIPC, DeepSeek said it used Beijing Volcano Engine Technology’s services to improve the security and user experience of its app, but later blocked the transfer of AI prompt information from April 10.

OpenAI calls for U.S. DeepSeek ban

DeepSeek and ByteDance did not immediately respond to inquiries from CNBC. 

The Hangzhou-based AI startup took the world by storm in January when it unveiled its R1 reasoning model, rivaling the performance of Western competitors despite the company’s claims that it was trained for relatively low costs and with less advanced hardware. 

However, the app’s rising popularity quickly triggered national security and data concerns outside China due to Beijing’s requirement for domestic firms to share data with the PRC. Cybersecurity experts have also flagged data vulnerabilities in the app and voiced concerns about the company’s privacy policy. 

PIPC on Thursday said it had issued a corrective recommendation to DeepSeek, which includes requests to immediately destroy AI prompt information transferred to the Chinese company in question and to set up legal protocols for transferring personal information overseas.

When the data protection authority announced the removal of DeepSeek from local app stores, it signaled that the app would become available again once the company implemented the necessary updates to comply with local data protection policy.

That investigation followed reports that some South Korean government agencies had banned employees from using DeepSeek on work devices. Other global government departments, including in Taiwan, Australia, and the U.S., have reportedly instituted similar bans.

Continue Reading

Technology

Adobe to launch mobile app for AI image generation tool as OpenAI steps up rivalry

Published

on

By

Adobe to launch mobile app for AI image generation tool as OpenAI steps up rivalry

Adobe’s new artificial intelligence image models, Firefly Image Model 4 and Firefly Image Model 4 Ultra, can generate hyper-realistic pictures in response to user prompts.

Adobe

LONDON — Adobe plans to launch a mobile version of its artificial intelligence image generation tool Firefly, stepping up a challenge to OpenAI as the Microsoft-backed startup advances its efforts on visual applications for the technology.

The design software giant said Thursday at its MAX creativity conference in London that it will release Firefly on both iOS and Android “soon,” without giving a specific date.

“Creative people think on the go,” Alexandru Costin, vice president of Adobe Firefly, told CNBC in an interview. “One of the visions we have is for the Firefly mobile application to become a creative partner that sits with you all the time.”

Costin said that one way creatives could use its upcoming mobile app was to ask it to sketch up some ideas about an ad campaign while commuting to the office, so that by the time they arrive at work they’ve got a mood board to help them develop their thinking.

Adobe also announced the launch of its latest AI models, Firefly Image Model 4 and Firefly Image Model 4 Ultra, and said its new Firefly Video Model for video generation is now generally available.

The company said the new systems are capable of generating hyper-realistic pictures and videos in response to textual prompts in a “commercially safe” way, blocking the inclusion of any intellectual property.

Competition from OpenAI

It marks Adobe’s latest push to incorporate AI into its creative tool suite and comes as the company is increasingly facing competition from well-funded AI firms such as OpenAI and Runway.

Last month, OpenAI released a native image generation feature that went viral online for its ability to produce anime images in the style of animation studio Studio Ghibli and recreate people as toy dolls.

The tool saw such huge levels of demand that OpenAI boss Sam Altman warned it was melting the company’s GPUs (graphics processing units). “It’s super fun seeing people love images in ChatGPT. But our GPUs are melting,” Altman said on March 27.

While Adobe’s Costin conceded that the competitive environment is heating up, he said the company isn’t shying away from partnering with the competition. For example, Adobe has partnered up with the likes of OpenAI, Google and Runway to add their AI image generation tools to Firefly.

“Competition is great,” Costin told CNBC. “We think there will be models with different personalities and capabilities.”

Continue Reading

Technology

British fintech Revolut tops $1 billion in profit as revenue jumps 72%

Published

on

By

British fintech Revolut tops  billion in profit as revenue jumps 72%

Revolut CEO Nikolay Storonsky at the Web Summit in Lisbon, Portugal, Nov. 7, 2019.

Pedro Nunes | Reuters

LONDON — British fintech firm Revolut on Thursday announced it topped $1 billion in annual profit for the first time, a major milestone for the company as it readies the launch of its U.K. bank later this year.

Revolut, which offers a range of banking and financial services via an app, said that net profit for the year ending Dec. 31, 2024, totaled £1.1 billion ($1.5 billion), up 149% year over year. Revenues at the company increased 72% year on year to £3.1 billion, driven by growth across different revenue streams.

Revolut’s wealth unit — which includes its stock-trading business — saw outsized growth, with revenue surging 298% to £506 million, while subscriptions turnover jumped 74% to £423 million.

Revolut also saw significant growth in its loan book, which grew 86% to £979 million. Coupled with a jump in customer deposits, this contributed to a 58% increase in interest income, which totaled £790 million.

UK bank rollout

Revolut’s financial milestone arrives at a critical time for the almost decade-old-firm. The digital banking unicorn has been preparing a transition to becoming a fully operational bank in the U.K. after securing a banking license last summer.

It was granted a banking license with restrictions in July 2024 from the U.K.’s Prudential Regulation Authority, bringing an end to a lengthy application process that began back in 2021.

The restricted license means that Revolut is now in the “mobilization” stage, where it is focusing on building out its banking operations and infrastructure in the run-up to a full launch. The period typically lasts about 12 months.

Revolut is still awaiting approval from regulators to transfer all 11 million of its U.K. users to a new banking entity this summer. Once fully up and running, the firm will be able to begin offering loans, overdrafts and mortgages, opening up the path to new income streams.

‘Customers trust banks’

Victor Stinga, Revolut’s chief financial officer, told CNBC on Thursday that the company’s aim is to formally launch its U.K. bank later this year.

“As you can imagine, at this scale, it’s a thorough process, and we just pay a lot of attention to it,” Stinga said. “We work very closely on a close contact with the PRA [Prudential Regulation Authority] and the FCA [Financial Conduct Authority] on it. We feel like we’re making great progress on it.”

Stinga said that a big advantage of becoming a bank in the U.K. is ability to start accepting deposits protected by government guarantees. Licensed banks are covered by the Financial Services Compensation Scheme, which means their customers can claim up to £85,000 if a lender goes out of business.

“Customers trust banks, so it means customers on this transition will use Revolut as a primary bank account,” Stinga said.

Lending is arguably “the biggest roadmap item that this unlocks,” Revolut’s CFO said, adding that the firm is looking at launching credit cards and personal loans, similar to the products it already offers in the European Union under a separate EU banking license.

Francesca Carlesi, Revolut’s U.K. boss, previously told the Wall Street Journal that Revolut views its journey to becoming a U.K. bank as a crucial step in its global expansion and eventual IPO. “My main strategic focus is making Revolut the primary bank for everybody in the U.K.,” she told the WSJ.

It has a steep hill to climb — rivals Monzo and Starling have had a lengthy head start on Revolut. Monzo obtained its full banking license in 2017, while Starling was granted its own permit in 2016.

Continue Reading

Trending