AI models keep getting smarter – but how smart can they get? – SiliconANGLE

As Nvidia became the world’s most valuable company this week, if only for a few days, it’s apparent that the race to develop even better generative artificial intelligence models isn’t slowing down.

This week we saw Anthropic provide a leap in its Claude large language model, along with news that OpenAI co-founder Ilya Sutskever’s new startup SSI is aiming to create a “safe superintelligence.” and OpenAI is buying real-time database provider Rockset. And investors keep ponying up to keep the race going, pouring $400 million into Poolhouse — who? — and eagerly anticipating a possible initial public offering by Nvidia AI chip rival Cerebras Systems.

Still, there’s a question of how much LLMs can advance in the near term, at least. According to benchmarks from Sierra Technologies, the startup from former Salesforce co-CEO Bret Taylor and former Google exec Clay Bavor, most LLMs fail at more complex tasks. Moreover, Wired did an investigation of how the popular AI search service Perplexity works, and it doesn’t look good.

Back in the mainstream enterprise, Hewlett Packard Enterprise CEO Antonio Neri at the company’s Discover event this week talked up hybrid cloud and edge as the key to AI, with a boost by Nvidia CEO Jensen Huang. But HPE’s proposed acquisition of Juniper hit a regulatory snag.

Meanwhile on the regulatory front, the Justice Department sued Adobe over its subscription policies and the Commerce Department banned Russia-based Kaspersky Labs from selling its cybersecurity software in the U.S. And social media may get slapped with tobacco-style warnings, not to mention New York state Gov. Kathy Hochul signed a bill intended to counteract social media harms.

This and other news will be discussed in depth on John Furrier’s and Dave Vellante’s weekly podcast theCUBE Pod, out this afternoon on YouTube. Also, don’t miss Vellante’s weekly deep dive, Breaking Analysis, due out this weekend.

Here’s the week’s news and analysis from SiliconANGLE and beyond:

AI and data: No slowdown in model advances

Nvidia surpasses Microsoft to become the world’s most valuable company (at least until Thursday when its stock dropped 3.5%)

OpenAI acquires enterprise information retrieval startup Rockset

Anthropic launches Claude 3.5 Sonnet to raise bar for model intelligence in coding and visual processing

OpenAI co-founder Ilya Sutskever launches startup to develop ‘safe superintelligence’

Bessemer’s latest State of Cloud report predicts AI’s world domination

This isn’t a good look, at the least, for Perplexity. But I think the real problem, even apart from the scraping, is not just Perplexity but the fact that generative AI’s “making stuff up” is an inherent feature that’s also a bug for some of the use cases it’s being applied to — not something that inevitably gets solved with more data: Perplexity Is a bullshit machine, says Wired

Bret Taylor’s and Clay Bavor’s startup claims to do better: AI startup Sierra’s new benchmark shows most LLMs fail at more complex tasks

Report: Apple’s generative AI ambitions likely to be delayed in China And in Europe: Apple delays release of Apple Intelligence, two other new AI features in the EU

Money matters

AI chipmaker Cerebras Systems, a competitor to Nvidia, reportedly files for an IPO

AI coding assistant startup Poolside reportedly raising $400M

Genspark reels in $60M for its AI-powered search engine

Daydream raises $50M to change online shopping using personalized AI discovery

Tinybird reels in $30M for its real-time data platform

Decagon raises $35M to transform customer service with more ‘humanlike’ AI agents

AI-powered e-commerce accounting startup Finaloop raises $35M

CuspAI raises $30M for AI-designed materials discovery and evaluation on demand

Multi-model database startup SurrealDB raises $20M and announces cloud beta access

Substrate reels in $8M to simplify AI application development

Materia launches with $6.3M to provide generative AI assistance for public accounting firms

More new models and services

Runway debuts new Gen-3 Alpha model for generating videos

Meta’s Fundamental AI Research team releases new AI models for audio generation, text-to-vision and watermarking

IBM’s generative AI-powered application automation tool Concert is now generally available

Oracle’s APEX now supports natural language prompts to ease app development

McKinsey offering aims to bridge the gap from AI prototypes to production

Adobe integrates more generative AI features into Acrobat

Researchers introduce OpenVLA, an open-source generalist AI model for robotics tasks

Finbourne raises £55M to break down financial data silos with AI

LogicMonitor says new AI assistant can reduce IT operations alerts by 95%

NICE customers provide best practices for adopting AI in the contact center

HPE charges hard into AI

News and analysis from a surprisingly energetic HPE Discover. The AI tide is truly lifting a lot of enterprise boats, especially those such as HPE that are joined at the hip with Nvidia:

Elsewhere around the enterprise

Governments crack down on tech:

Amazon’s first labor union partners with the Teamsters Meanwhile, Amazon was fined by California for failure to disclose worker quotas

Amazon to spend €10B on expanding its cloud and logistics operations in Germany

Nvidia reportedly acquires incident automation startup Shoreline for $100M

SUSE acquires Kubernetes observability startup StackState

Pure Storage adds AI features to improve performance and detect attacks

Oracle Autonomous Database is now an Azure service

Aryaka debuts dedicated network for accelerating generative AI workloads

FinOps Foundation debuts unified billing model for cloud computing

Collaborative developer tool Multiplayer launches with IT architecture auto-documentation feature

D-Wave announces new hybrid quantum solver for commercial applications supporting 2M variables

Salesforce launches Life Sciences Cloud for organizing clinical trials and patient outreach

Big bucks keep flowing into cybersecurity

Huntress raises $150M at $1.5B+ valuation for its managed cybersecurity platform

Cyber resilience startup Semperis raises $125M for enterprise identity protection expansion

Post-quantum cryptography startup PQShield raises $37M

Israeli startup Aim Security raises $18M to bolster generative AI security

Entro Security raises $18M to expand global operations in identity and secrets security

Attacks

Ransom demands issued to Snowflake users amid alleged third-party contractor breach

Hacking group Qilin leaks data stolen from UK medical testing provider Synnovis

Cyberattack on CDK Global disrupts car sales in North America

Hackers demand $50M ransom payment from UK lab provider following hospital disruption

Suspected key member of Scattered Spider cybercrime group arrested in Spain

AMD investigates alleged data theft as stolen data appears for sale on BreachForums

Barracuda report: AI-enhanced phishing drives surge in email security threats

LockBit resurgence sees ransomware attacks reach record high in May

New products

Zeus Kerravala sums up Zscaler’s recent conference: Zscaler Zenith Live ‘24 touts breach predictions, zero trust and expanded partnership with Google And Shelly Kramer analyzes the Google partnership along with her own event roundup: Zscaler-Google strategic alliance forged to enhance enterprise security and access

EchoMark says it has the definitive answer to data exfiltration

Cato Networks introduces new partner program to simplify managed SASE services

ZeroFox unveils new mobile app for real-time physical security intelligence

Next DLP launches Secure Data Flow to prevent data theft and enhance security

Elsewhere in tech

Waabi secures $200M for AI-powered autonomous trucks, targets 2025 launch

GrayMatter Robotics raises $45M to automate tricky manufacturing operations

French quantum computing startup C12 raises €18M to stabilize qubits with carbon nanotubes

Varjo unveils Teleport, a service that lets smartphones turn real-world places into VR scenes

Niantic launches Studio for web-based immersive 3D and mixed reality creation

Comings and goings

Stability AI appoints new CEO, per The Information: Prem Akkaraju, former CEO of visual effects company Weta Digital. He’s part of a group, also including former Facebook President Sean Parker, that’s reportedly providing a cash infusion at a much lower valuation.

Developer cloud provider DigitalOcean appoints AWS AI, machine learning and data infrastructure veteran (and CUBE alum) Bratin Saha chief product and technology officer

What’s next

Conference season slows down, but there are a few:

June 23-27: Design Automation Conference, San Francisco

June 25-27: AI Engineer World’s Fair, San Francisco: new one to me but a pretty decent lineup of speakers

June 26-27: Figma’s Config, San Francisco

Earnings: Wednesday, June 26: Micron

Image: SiliconANGLE/Ideogram