HARDWARE

IBM partners with chip firm Groq

As AI developers grapple with rising demand for compute, alternatives to traditional GPUs are emerging. 

On Monday, IBM announced a partnership with Groq, a startup that produces chips specifically designed for inference, to bring its technology to IBM’s agent orchestrator platform. The partnership will allow customers of IBM’s watsonx Orchestrate to access the power of Groq’s “Language Processing Units,” or LPUs, for faster, lower latency and cheaper inference tasks. 

Additionally, IBM’s Granite family of models will be supported on GroqCloud, the company’s inference platform. 

Stuart Pitts, head of global product marketing at Groq, told The Deep View that IBM brings “deep, rich experience” in critical enterprise business verticals, while Groq brings “clear product differentiation from an inference perspective.” 

Though the financial details of the partnership weren’t revealed, the deal marks another win for Groq, which recently raised $750 million in a series E round at a valuation of $6.9 billion. The company is emerging as a challenger to traditional chip providers like Nvidia as it offers inference capabilities with its LPUs and developers seek alternatives. 

Pitts said that Groq customers often face three issues with traditional GPUs: Speed, cost and reliability. That’s because the AI landscape is shifting from training models – which is the GPU’s purpose – to using them for inference, he said. He compared Groq’s LPUs to a “purpose-built assembly line” specifically for inference tasks. 

“A GPU is a fantastic device for the purposes in which it was designed,” Pitts said. “But a GPU was never designed for this AI-first world.” 

The success of a chip firm like Groq is a sign of the times. Companies, broadly, are looking for ways to do AI cheaper, and the key to that is purpose-built architecture. It’s why neocloud firms like Nscale and CoreWeave have skyrocketed in popularity. However, taking on a legacy isn’t easy, especially when that legacy is the chip behemoth Nvidia. Even other GPU firms like AMD struggle to claw market share from the likes of Nvidia. But with the landscape and demands of AI constantly shifting, the firms like Groq are starting to share the spotlight.

TOGETHER WITH IBM

Data in the driver’s seat

With IBM, Scuderia Ferrari HP focused on data to fuel strategy, precision, and passion. On the track and beyond, data became an engine of innovation for fans.

Ferrari’s fan app transformed raw race data into immersive, real-time storytelling, including:

• AI-generated race summaries from live and historical data

• Scalable hybrid cloud infrastructure for enhanced speed and agility

• Real-time personalization based on fan behavior

Every lap, corner and millisecond becomes a window into the data-driven decisions that brings fans closer to the team like never before.

Know more

CULTURE

OpenAI, MLK controversy signals deepfake dilemma

AI has made it easier than ever to rewrite history.

On Thursday, OpenAI paused Sora’s ability to generate videos using the likeness of Martin Luther King Jr. at the request of his estate. The move follows some users generating “disrespectful depictions” of his image, the company said in a post on X. 

“While there are strong free speech interests in depicting historical figures, OpenAI believes public figures and their families should ultimately have control over how their likeness is used,” OpenAI said in the post. 

The company noted that it is also in the process of strengthening its guardrails for generating content surrounding historical figures. Authorized representatives and estate owners can now request that their likeness be restricted from use by Sora, OpenAI said. 

Dr. King’s estate joins Zelda Williams, the daughter of Robin Williams, in calling on users to stop generating and sending her AI videos of her father. Google faced similar backlash last year for its models generating historically inaccurate images of the Founding Fathers.

The news highlights growing controversy surrounding AI-generated depictions of famous figures, stoking both copyright concerns and issues surrounding historical accuracy, Ben Colman, CEO and co-founder of Reality Defender, told The Deep View.

“From day one, Sora let anyone with an access code make anything of anyone regardless of permission, rights ownership, and beyond,” said Colman. “Many of these depictions were off-brand at best and horribly insensitive at worst.” 

And this dilemma extends beyond Sora, Colman said. Though OpenAI’s platform has been a popular choice, practically every major tech firm is throwing their hat into the video generation ring. 

“The impact of this ability to do this anywhere with nary a check on major content distributors (read: social media platforms) is to repeatedly meme false AI slop-led history into the conscious of others,” Colman said.

TOGETHER WITH WISPR FLOW

Your Keyboard Is Killing You

Unless you’re the Guinness World Record Holder for “Most Words Typed Per Minute”, we all talk faster than we can tap out a sentence – and that means your keyboard is slowing you down. Well, Wispr Flow is here to fix that (and give your hands a much-needed break). 

This voice-powered AI app is 4x quicker than typing, making it easier than ever to express yourself clearly, quickly, and without any issues. Wispr Flow uses AI to auto-edit your dictation on the fly, even learning your most-used phrases and acronyms so you don’t have to go back and fix them yourself. Plus, it works inside your existing apps (think Slack, Notion, and more) with no setup.

You talk, it types, everyone wins. Try Wispr Flow right here.

CONSUMER

Meta reins in teen AI use

Meta is pulling in the reins on its AI chatbots. 

The social media firm announced on Friday that it’s introducing parental controls for its AI interactions with teen accounts. Parents can now get insights on their teens’ AI use within Meta, block interactions with specific AI characters or turn off teens’ access to one-on-one chats with these models entirely. 

This update adds to Meta’s existing AI protections, which automatically apply PG-13 responses to teen accounts and restrict discussion of self-harm, suicide or disordered eating. 

“AI is evolving rapidly, which means we are going to need to constantly adapt and strengthen our protections for teens, while listening and responding to concerns parents have about this new technology,” Instagram head Adam Mosseri and Meta Chief AI officer Alexandr Wang said in a blog post.

Meta faced backlash in August for allowing its AI chatbots to have romantic interactions with young users. These restrictions come amid growing concern over the ways that minors interact with chatbots, as many turn to the tech for judgment-free emotional support without comprehending the risks.

OpenAI implemented similar controls in late September, allowing parents to set quiet hours for ChatGPT teen accounts, opt out of saved memories, automatically limit graphic or inappropriate content and inform parents of acute distress or imminent danger. 

These firms may also be seeking to get ahead of regulatory scrutiny following California Gov. Gavin Newsom’s signing of SB 243 earlier this week, which regulates AI companions by requiring developers to implement age verifications and protocols addressing discussions of self-harm.

LINKS

• Microsoft, OpenAI to train teachers on AI for classroom use

• Apple, F1 reach a five-year media deal worth $140 million annually

• ChatGPT downloads, daily user growth is slowing, data shows

• Revolut nears the end of $3 billion funding round, valuing it at $75 billion

• Stellantis, Pony.ai team up to build robotaxis in Europe 

• Meta AI can now suggest edits to photos in your camera roll

• Radiant: A free AI assistant from Supernormal that turns your meetings into ready-to-send follow-up emails. Meetings in, momentum out. (sponsored)

• Manus 1.5: The newest model from Manus, now with full-stack web application development. 

• Meshy AI: The latest release from Fal, offering accurate AI-powered 3D reconstruction. 

• Capcut AI Design: A “flat image design tool” from Capcut, live now on desktop and web with 10 days of free use. 

• Leadbay: Discovers and qualifies business leads in data-scarce and fragmented industries.

• Together AI: GPU Cluster Resource Scheduling and Optimization Engineer

• OpenAI: Lead Research Engineer / Scientist, Cyber

• Anthropic: Research Scientist/Engineer, Honesty

• Amazon: Applied Scientist III, Next Gen Developer Experience

GAMES

A QUICK POLL BEFORE YOU GO