- AI For All
- Posts
- The Empire Strikes Back
The Empire Strikes Back
Gemini 2.0 and Sora's Failure
Hello readers,
Welcome to the AI For All newsletter! Google unveiled Gemini 2.0 with agentic capabilities and is making the chips used to train it generally available. Meanwhile, OpenAI suffered a disastrous Sora launch. What happened? Let’s find out!
The Empire Strikes Back
Google released Gemini 2.0 with agentic capabilities, or to be more accurate, an experimental version of Gemini 2.0 Flash (their smaller model) and prototypes of their agents — so not the real deal. I’ve given up trying to understand what all the versions of Gemini are and where to access them. The announcement opened with a message from Sundar Pichai that claimed, “Information is at the core of human progress.” Pichai should specify what kind of information, such as factual information, which AI struggles with.
Gemini 2.0 Flash will be natively equipped to take actions on a user’s behalf, starting in the browser with Project Mariner. Meanwhile, developers will be getting Jules to assist with coding tasks, and gamers will be getting virtual gaming companions that can “help you navigate the virtual world of video games” and “offer up suggestions for what to do next” like you’re a three-year-old. Isn’t the whole point (and fun) of a video game to explore and discover everything for yourself? While all of these agents are considered experimental, I’m sure they’ll be rushed out soon for all of us to be frustrated by.
When it comes to agents, forgive me for not awkwardly jumping up and down with glee like Elon Musk at a Trump rally, but the way the AI industry has collectively leaped to agents (before even getting plain LLMs working reliably) indicates to me a need to prolong a hype cycle that is losing steam. At this year’s DealBook Summit, Pichai even admitted that AI progress will get harder because “the low-hanging fruit is gone.”
Google is making their sixth-generation Trillium TPUs (used to train Gemini 2.0) generally available to Google Cloud customers. However, you may want to wait until Google’s new quantum chip, Willow, becomes widely available in thirty years. Though seriously, if Google can harness quantum for its AI efforts, then something may or may not happen. I know, these are the kind of stunning insights you come here for.
In other news, OpenAI finally released Sora, its AI video generator, only to quickly suspend new account creation because, according to the CEO, they “significantly underestimated demand.” As of writing this, new account creation is back (for now). Sora’s reception has been mixed, with so-called “AI filmmakers” reporting inconsistent and unrealistic results — gee, you think? The showcase on sora.com consists of the typical horrific sludge we’ve gotten used to from every other video generator. And to think, OpenAI wants this grotesque thing to function as a world model (it won’t).
We’re six days into the 12 Days of OpenAI series, and five out of the six announcements are things we already knew about, such as advanced voice with video, which along with Sora, will only lose OpenAI more money. It is unclear what the point of generative video is other than to empower the worst among us — scammers, bad actors, and talentless hacks. The concept of “democratizing creativity” is nonsensical. Generative AI is exactly what one uses in place of creativity. Efficiency at the expense of quality is not worth it.
🔥 Rapid Fire
Topical: the possibilities and perils of AI in the health insurance industry
Meta releases Llama 3.3 70B parameter model for lower cost development
Google rolls out Deep Research tool to Gemini Advanced subscribers
Cognition makes AI coding assistant Devin generally available for teams
Google announces Android XR in partnership with Samsung and Qualcomm
Apple reportedly collaborating with Broadcom to develop AI server chip
Google partners with cleantech to co-locate data centers and clean power
Exxon announces plans to build fossil fuel plant to power AI data centers
Amazon opens new AI lab in San Francisco focused on long-term research
NVIDIA shares fall after China opens investigation into monopoly violation
Adobe shares fall as investors doubt company’s competitiveness in AI
Pentagon sunsets generative AI task force, launches rapid capabilities cell
Russia teams up with BRICS to form AI alliance amid Western dominance
Need a personal assistant? We do too, that’s why we use AI.
Ready to embrace a new era of task delegation?
HubSpot’s highly anticipated AI Task Delegation Playbook is your key to supercharging your productivity and saving precious time.
Learn how to integrate AI into your own processes, allowing you to optimize your time and resources, while maximizing your output with ease.
📖 What We’re Reading
“As the generative AI map takes shape, the US and China are asserting their dominance. But something interesting is happening in parallel. A small group of countries—the GenAI middle powers—is emerging, each with its own distinct strengths that may enable it to compete on a regional and even global scale as a supplier of the technology. The implications for companies are significant.”
💻️ AI Tools and Platforms
Zed → Code editor with AI integration
Devin → Collaborative AI teammate for coding
Contextual AI → Build enterprise RAG applications
Flowith → AI for deep work and productivity
Midship → All-in-one document intelligence