☕️ Elon Musk's xAI Catches Up To Industry Leaders With Grok-2

Elon Musk's xAI has finally caught up with the rest of the pack with Grok-2, meeting Google, OpenAI, and Anthropic at their level, but not without some classic Elon controversies. Other key highlights include:

• Google releases its Gemini Live voice mode to compete with OpenAI's Advanced Voice Mode

• Imagen 3, Google's most capable text-to-image model, silently becomes available to all US users

• Silicon Valley's criticism pushes California to make changes to SB 1047, its AI disaster prevention bill

Join us at AI Tangle as we untangle this week's happenings in AI!

xAI's Grok-2 beta version pops up on the LMSYS AI model leaderboard on equal footing with industry leaders

In the middle of the week, users on LMSYS noticed a curious model under the alias of "sus-column-r" pop up on the AI model leaderboards, gaining rapid traction and meeting models like Claude 3.5 Sonnet and GPT-4o on their level. ​A little while later​, xAI claimed responsibility, stating that it was a very early beta version of Grok-2 and that it would be released with a Grok-2 mini variant. xAI also adds that developers should expect access to the company's frontier models through its enterprise API later this month.

​

​​And it comes with a ridiculously uncensored image generator​​

Along with the Grok-2 debut news, however, came an interesting feature to xAI's Grok chatbot - a completely unrestrained AI image generator powered by the FLUX.1 model by Black Forest Labs. X Premium and Premium+ members have been using Grok's image generation feature to post anything from ​Donald Trump flying a plane into the World Trade Center buildings​ to female celebrities in lingerie on Elon Musk's social media platform. The loose guardrails on Grok's image generation feature line up with Musk's general distaste towards standard AI and social media safety conventions and his unconventional approach to freedom of speech.

​CustomGPT.ai: Driving Revenue with Reliable AI Agents​

CustomGPT.ai has redefined AI reliability, outperforming industry giants with a 10% lower hallucination rate, 13% higher accuracy, and 34% faster responses.

As the leading no-code platform, we empower businesses to harness this superior AI technology, creating reliable AI agents that not only boost efficiency but directly impact revenue growth.

• Benchmark Superiority: Lower hallucinations, higher accuracy, faster responses.

• No-Code Platform: Rapid deployment of reliable AI agents.

• Versatile Applications: Customer Support, Sales, Product, and beyond.

Google gives its Gemini AI assistant a voice mode along with more features across its apps

Google's recent Pixel 9 live event brought many new developments to light, including its Gemini Live voice chat mode for Gemini Advanced subscribers, which was announced during its I/O developer conference earlier this year. With 10 voices to choose from, Google describes Gemini Live as a "free-flowing" conversationalist, similar to OpenAI's recent Advanced Voice Mode. Gemini will additionally be getting extensions to some apps like Keeps, Utilities, Tasks, and YouTube Music, along with screen-aware context. Gemini Live is currently only available on Android and in English, but an iOS release and more languages are expected "in the coming weeks."

​

Google silently makes its most advanced image generation model, Imagen 3, available to US users

Aside from the Pixel 9 live event, however, Google had also quietly made its most capable image generation model, Imagen 3, available through the ImageFX platform to US users. The company's research paper highlights and details its developments and the advances the team at Google made, but the reception has been somewhat mixed. Though praises get sung about its improved texture and word recognition capabilities, many users are frustrated with the model's more restrictive content filters and censorship.

​

Silicon Valley and Anthropic push California to make amendments to its AI disaster prevention bill

California's SB 1047 bill, designed to prevent AI disasters, recently passed the Appropriations Committee after amendments following criticism from Silicon Valley, including Anthropic, now heading to the Assembly floor for a final vote. The bill now uses softer language, adds protection to open-source fine-tuned models, and limits the attorney general's ability to sue AI companies, allowing legal action only after a disastrous event. The creation of the Frontier Model Division agency is also being axed but retains a regulatory board within the existing Government Operations Agency, which increased from five people to nine.

​

Opera releases its AI-focused Opera One web browser on iOS devices

Opera One, the AI-focused web browser launched a year ago by, you guessed it, Opera, is now available for iOS following a 63% growth in Europe after the Digital Markets Act. It features a full-screen interface and mobile-specific enhancements, including a movable search bar and predictive search chips. The browser comes with Opera's Aria AI assistant with voice input and image generation powered by Google's Imagen2. As for security features, Opera One includes a built-in ad blocker and a VPN.

​

AI pioneer Fei-Fei Li's AI startup World Labs bags $100 million in funding and achieves unicorn status

Stealth startup World Labs, founded by AI pioneer Fei-Fei Li, recently achieved unicorn status with a $100 million funding round led by NEA, putting the company valuation at just over $1 billion barely four months after its founding in April. The startup is focused on developing AI models that accurately estimate the three-dimensional physicality of real-world objects, enabling detailed digital replicas with minimal data collection. Investors, including Andreessen Horowitz and Radical Ventures, are betting big on Li's vision, which could have major implications for gaming and robotics.

​

EliseAI raises $75 million in Series D funding to help property managers handle renters

Also joining World Labs on the list of unicorns is EliseAI, the leading conversational AI platform for the housing industry, recently announced that it had raised $75 million in a Series D funding round, boosting its valuation to over $1 billion. Led by Sapphire Ventures, EliseAI aims to expand its business into healthcare to handle administrative tasks. Company CEO and co-founder Minna Song said that the company looks forward to partnering with the Sapphire Ventures team to handle essential industries like housing and healthcare.

​Tely - Struggling to attract niche B2B leads? Tely AI's autonomous AI agent generates expert content, drives organic traffic, and captures leads in just three weeks. ​Get quality content and fast results.​*

​Salesify - Level your sales game with Salesify, a powerful AI-powered sales coach that analyzes pain points and provides insights on lost (or won) deals to improve how you run sales.

​Fotor - Fotor is an AI-enhanced online photo editor with features to enhance and edit photos effortlessly - experience simple online photo editing at your fingertips.

​Recap - Recap by Fabric is a personal "read it later" newsletter of everything you've saved, created, or captured recently, sent directly to you with a convenient AI summary to help you reflect and review.

* indicates a promoted tool

To Change Filmmaking Beyond Recognition (4-min read)

Generative AI makes for neat demo highlight reels and could theoretically revolutionize filmmaking, but experts debate its current limitations and future potential. Will AI create a new genre or fall short of expectations?

Are NPUs The Next Big Deal For AI? (15-min watch)

Tech giants like Microsoft, Google, Apple, Qualcomm, Intel, and AMD are all pushing NPUs (Neural Processing Units) or neural engines in their latest devices to handle AI workloads. Tech analytics channel TechAltar aims to determine whether these chips deserve the attention they're getting and whether they're effective.