The generative AI boom is hitting some turbulence; the BBC mulls opening its archives to AI companies; AI spots tiny cancers missed by doctors; UN adopts first AI resolution

Microsoft and Inflection AI in "licensing deal"; transformer pioneers reunite at GTC; open source attempts to crack OpenAI's dominance; Perplexity takes on Google on search

While everyone else was following the deluge of news coming out of GTC this week, two very different stories caught my eye: first, Reuters reported that Cohere was seeking to raise $500 million at a valuation of $5 billion. This comes less than a year after the startup raised $270 million as part of its Series C round. The reason why I found the report interesting however was another figure — Cohere’s generated about $22 million in annualized revenue this month, up from $13 million at the end of last year.

Then came another interesting article from The Information: as part of a deal between Microsoft and Inflection AI (which generated quite a stir because Mustafa Suleyman, Inflection and DeepMind’s co-founder, joined Microsoft to lead Copilot), the tech giant agreed to pay the startup approximately $650 million to ensure “a good outcome” for Inflection’s investors.

When taken together with other recent stories of staff departures and funding challenges, it is becoming increasingly clear that the generative AI boom is hitting some turbulence, with the AI startup landscape looking a lot riskier than it did just six months ago. Many of these companies are burning through cash at an unsustainable rate while operating in an increasingly crowded field driven by the commoditization of large language models that threaten to further squeeze many AI upstarts before they ever reach escape velocity.

The generative AI goldrush of 2023 was fueled by large language models but developing and operating these compute-hungry systems is phenomenally expensive. For example, the estimated compute costs for training a model akin to Meta’s Llama 2 was upward of $20 million; Mistral recently revealed that they required 15,000 H1 GPUs to develop their latest Mistral Large model; and Perplexity’s CEO shared that some researchers wouldn’t join his company unless he had 10,000 GPUs to spare. For most startups, these economics are unsustainable long-term, especially as some large tech companies release open source models that are similar in performance.

CB Insights’ most promising startups of 2023 list

We're therefore rapidly heading for a scenario where large language models become a resource that any company or developer can access cheaply via cloud APIs or even download from Hugging Face for free. That undercuts one of the main competitive advantages for the current crop of generative AI startups.

While many companies in the space followed an "AI-first" model (developing large language models before figuring out commercial applications), they're now racing to find a sustainable business ahead of their funds running dry. Some players are pivoting away from consumer applications and into the enterprise space. Others are doubling down to try and pull ahead of the pack.

But as the AI mania among VCs is cooling and many of the experiments that went viral on social media last year fail to reach deployment at scale in 2024, startups will need to demonstrate truly differentiated value to justify the operational costs of continuing to push the envelope on models. In an ironic twist, many generative AI startups face an existential threat of their own creation: by hastening the development of large, powerful language models capable of automating so much, they may have inadvertently rushed their own demise unless they pivot quickly.

The companies that survive in 2025 may ultimately look very different than the ones earning all the hype on X today.

And now, here are this week’s news:

❤️Computer loves

Our top news picks for the week - your essential reading from the world of AI

• MIT Technology Review: Google DeepMind’s new AI assistant helps elite soccer coaches get even better

• FT: BBC develops AI plans and talks to Big Tech over archives access

• VentureBeat: ‘Attention is All You Need’ creators look beyond Transformers for AI at Nvidia GTC: ‘The world needs something better’

• WSJ: Open-Source Companies Are Sharing Their AI Free. Can They Crack OpenAI’s Dominance?

• TechCrunch: Why is AI so bad at spelling? Because image generators aren’t actually reading text

• Wired: Perplexity's Founder Was Inspired by Sundar Pichai. Now They’re Competing to Reinvent Search

• AP: How bracketologists are using artificial intelligence this March Madness

• VentureBeat: The first 'Fairly Trained' AI large language model is here

• Bloomberg: OpenAI Sprinting to Keep Up With Startups on AI-Generated Video

• Wired: 8 Google Employees Invented Modern AI. Here’s the Inside Story

• Reuters: UN adopts first global artificial intelligence resolution

• BBC: NHS AI test spots tiny cancers missed by doctors

⚙️Computer does

AI in the wild: how artificial intelligence is used across industry, from the internet, social media, and retail to transportation, healthcare, banking, and more

• The Telegraph: Can a portable ultrasound scanner really cut childbirth deaths across Africa?

• VentureBeat: King researchers talk about the results of using AI at GDC 2024

• Fortune: AI is upending eye exams for patients and providers—in a good way: ‘It’s been a godsend for us’ 

• Fortune: How AI could help make the IVF process easier

• TechCrunch: GitHub’s latest AI tool can automatically fix code vulnerabilities

• Business Insider: NBCUniversal is rolling out generative AI-driven ads and says they perform up to 49% better in tests

• The Verge: Microsoft’s first AI PCs are the Surface Pro 10 and Surface Laptop 6 for businesses

• Fortune: AI is about to change the homebuilding process, from start to finish

• The Verge: Adobe Substance 3D’s AI features can turn text into backgrounds and textures

• Fortune: How AI can make U.S. cities smarter, safer, and greener

• Reuters: Central banks use AI to assess climate-related risks

• BBC: Hospitals using AI to diagnose prostate cancer

• VentureBeat: Roblox introduces AI-powered tools to boost 3D content creation

• Business Insider: These sustainability-tech companies use AI to improve ocean health by keeping plastics out, boosting coral reefs, and monitoring fish farms

• Bloomberg: Google Researchers Use AI to Predict Floods Where There’s Little Data

• Business Insider: I visited the ExploreAI summit in Singapore and here are the highlights on how AI is helping businesses solve real-world problems.

🧑‍🎓Computer learns

Interesting trends and developments from various AI fields, companies and people

• WSJ: Apple Held Talks With China’s Baidu Over AI for Its Devices

• MIT Technology Review: This self-driving startup is using generative AI to predict traffic

• The Economist: Just how rich are businesses getting in the AI gold rush?

• The Atlantic: Elon Musk Just Added a Wrinkle to the AI Race

• Fortune: Dead by 2030—C-suite execs set deadline for embracing AI at scale

• Fast Company: This is Nvidia’s lesser-known plan to stay dominant in the AI chip business

• VentureBeat: Nvidia partners with Shutterstock, Getty Images on AI-generated 3D content

• The Information: The Sora Saga Has Only Just Started

• The Verge: Nvidia reveals Blackwell B200 GPU, the ‘world’s most powerful chip’ for AI

• TechCrunch: Nvidia’s Jensen Huang says AI hallucinations are solvable, artificial general intelligence is 5 years away

• The Verge: Why Figma CEO Dylan Field is optimistic about AI and the future of design

• VentureBeat: Nvidia enables high-res Omniverse Cloud imagery on Apple Vision Pro

• TechCrunch: Nvidia and Qualcomm join Open Source Robotics Alliance to support ROS development

• Business Insider: Elon Musk says AI could be smarter than all of humanity in 5 years. Meta's AI chief thinks otherwise.

• Business Insider: A Lady Gaga Google Search shows how AI is upending the world's most profitable online business. 'Site owners are terrified.'

• VentureBeat: What’s a NIM? Nvidia Inference Manager is new approach to gen AI model deployment that could change the industry

• VentureBeat: Nvidia announces Earth-2 digital twin to forecast planet’s climate change

• CNBC: It’s generally a mistake for large businesses to rush into AI, technology company says

• The Information: OpenAI’s Chatbot App Store Is Off to a Slow Start

• Fast Company: Thanks to generative AI, advertising’s business models will be reinvented

• The Verge: Ubisoft let me actually speak with its new AI-powered video game NPCs

• FT: Can AI make CEOs more relatable?

• CNBC: Why ‘open’ AI might be more marketing than reality

• Bloomberg: AI Driving US Electricity Growth Rate By 81%, NextEra Says

• Bloomberg: Apple Is in Talks to Let Google Gemini Power iPhone AI Features

• Business Insider: What creators were talking about at SXSW, from the big AI push to creating spaces for influencers of color

• ZDNet: YouPro lets me access every popular premium AI chatbot for $20/month - but there's a catch

• Wired: The Filmmaker Who Says AI Is Reparations

• Wired: Here's Proof You Can Train an AI Model Without Slurping Copyrighted Content

• Business Insider: OpenAI is expected to release a 'materially better' GPT-5 for its chatbot mid-year, sources say

• Bloomberg: Reddit COO Says Platform ‘Incredibly Important’ for Training AI

• New York Times: Saudi Arabia Plans $40 Billion Push Into Artificial Intelligence

• Bloomberg: Tencent to Boost Mideast Cloud Investments Amid Regional AI Push

• Business Insider: An OpenAI mystery: YouTube videos, Google throttling, and AI training data.

• Washington Post: The hidden secrets of a simple birdsong, explained

• CNN: Artist collaborates with AI to create immersive installation of human heartbeats in real time

• The Verge: Nvidia is using AI to turn game characters into chatbots

• The Verge: Google has a new head of Search — and she’s all in on AI

• MIT Technology Review: How AI taught Cassie the two-legged robot to run and jump

• Reuters: Meta does not expect new Nvidia chips to arrive until at least next year

• Reuters: Nvidia's new AI chip to be priced at over $30,000, CNBC reports

• VentureBeat: Google researchers unveil ‘VLOGGER’, an AI that can bring still photos to life

• VentureBeat: Musk’s Grok AI goes open source

• VentureBeat: Apple researchers achieve breakthroughs in multimodal AI as company ramps up investments

• CNN: Digital humans: the relatable face of artificial intelligence?

• CNN: Apple is getting serious about AI

• CNN: Why this math professor is putting actors in classrooms

• Business Insider: Microsoft hiring of Inflection staff came amid lack of consumer AI vision, and board concerns about OpenAI, insiders say

• Business Insider: You'll have to pay at least $30,000 to get your hands on Nvidia's new chip, CEO says

• Axios: Generative AI is a "big unlock" for Slack, CEO says

• Wired: Apple’s MM1 AI Model Shows a Sleeping Giant Is Waking Up

• VentureBeat: Exclusive: AWS, Accenture and Anthropic partner to accelerate enterprise AI adoption

• VentureBeat: Runway partners with Musixmatch, allowing musicians to generate AI lyrics videos

• VentureBeat: Nvidia unveils Latte3D to instantly generate 3D shapes from text

• VentureBeat: Snowflake teams up with Reka to add multimodal LLMs to data cloud

• VentureBeat: CIOs share how they are harnessing gen AI’s potential at Nvidia GTC

• VentureBeat: Google to IBM: How big tech giants are embracing Nvidia’s new hardware and software services

• WSJ: Nvidia Is Now Competing Mostly With Itself—and AI Fatigue

• WSJ: Apple Looks to External Partners to Boost AI Efforts

• Business Insider: Project Theia: Amazon's secret AI for online shopping inspiration

• TechCrunch: OpenAI's chatbot store is filling up with spam

• TechCrunch: ServiceNow is developing AI through mix of building, buying and partnering

• Independent: M&S to NHS: Former supermarket boss hired to help hospitals use AI

• Fortune: Sam Altman is over GPT-4: ‘I think it kind of sucks’

• FT: AI may not change your job, but it will transform government

• Forbes: Why $4.6 Billion Health Records Giant Epic Is Betting Big On Generative AI

• The Economist: A new generation of music-making algorithms is here

• Reuters: Synopsys says new tools help design cars, data centers faster

• The Economist: Can anything stop Nvidia’s Jensen Huang?

• Fortune: What Nvidia’s new Blackwell chip says about AI’s carbon footprint problem

• Fortune: Moderna started upskilling employees on AI 2 years ago. CHRO Tracey Franklin says they’ve learned some big lessons along the way

• Business Insider: Demand is skyrocketing for prompt engineers, one of the hottest roles in AI. But some say it won't last.

• Business Insider: It's hard to compete with Elon Musk when recruiting AI talent, CEO says

• Business Insider: Perplexity CEO slams Google's Gemini for creating flawed images: 'It's not a small bug'

• Business Insider: From the 'godfathers of AI' to newer people in the field: Here are 17 people you should know — and what they say about the possibilities and dangers of the technology.

• Business Insider: Former Salesforce exec says artificial general intelligence should be able to set its own goals, and that's still a long way off

• Axios: "We need aspirational AI," Girls Who Code founder Reshma Saujani says

• Business Insider: I tried the AI robot massage table coming to Equinox. It's impressive, but I still prefer a human.

• Business Insider: This company is using the AI that powers ChatGPT to help warehouse robots expect the unexpected

• Business Insider: I've made over $50,000 building AI tools for businesses. My main challenge: managing clients' expectations.

• Business Insider: Bill Gates says AI won't magically solve problems that humans aren't already good at

• Reuters: Nvidia expands ties with Chinese EV makers as auto AI race heats up

• VentureBeat: Smartling leads translation industry transformation with AI, driving 40% growth amid industry downturn

• VentureBeat: Stability AI brings a new dimension to video with Stable Video 3D

• VentureBeat: Nvidia shows off Project GR00T, a multimodal AI to power humanoids of the future

• VentureBeat: Nvidia partners with Shutterstock, Getty Images on AI-generated 3D content

• WSJ: Nvidia Plans to Price Newest AI Chips to Appeal to Wide Group of Users

• WSJ: For These Companies, It Feels Good to Be in Nvidia’s Orbit

• CNBC: Nvidia and Johnson & Johnson to develop new AI applications for surgery

• Bloomberg: Microsoft Hires DeepMind Co-Founder Suleyman to Run Consumer AI

• Bloomberg: Google Deepens AI Push Into Health Care With Fitbit, Screenings

• Bloomberg: Samsung Creates Lab to Research Chips for AI’s Next Phase

• Bloomberg: Novo Foundation, Nvidia Team Up to Build Danish AI Supercomputer