Meta Targeting OpenAI's GPT-4, Stability AI's New Groove, & Salesforce Announces AI Copilot
Meta is going full steam ahead on AI, Stability AI is charging for audio, and Salesforce finally announces their generative AI product
Meta's AI Ambitions: Targeting GPT-4, While Staying Open Sourced
Facebook parent Meta is reportedly cooking up a new AI model aiming to rival OpenAI's GPT-4 in power.
Just two months after releasing their "Llama 2" model, they're already setting their sights higher. This new project, still under wraps, is part of Zuckerberg's plan of asserting Meta's AI dominance after some perceived lagging. And while they previously partnered with Microsoft for Llama 2, this time, they're going solo, beefing up their data centers and hoarding Nvidia's top-tier chips.
Even if OpenAI & Anthropic develop better models by the time this is ready, Meta’s open-source will be a huge differentiator. Giving an open-source model like this to developers is like giving free steroids to Arnold Schwarzenegger when he just started weight training. An open-source LLM as powerful as GPT-4 helps developers launch applications that were unthinkable just a couple years ago, and they won’t be burdened with hefty licensing fees that could come from OpenAI or Anthropic.
Stability AI's New Groove: From Pictures to Pitch-Perfect Audio
Stability AI, previously making waves with its image generation tech Stable Diffusion, has now tuned into the audio realm with its latest offering, Stable Audio.
This tool lets users whip up short audio clips using plain text prompts. Built on similar AI techniques, it's like Stable Diffusion but for your ears. And if you're dreaming of recreating a Beatles' hit, just let it be. Their focus is on originality, not imitation from popular artists. In fact, the models haven’t even been trained on music from the Beatles.
Stability AI offers a free version, as well as Pro plan at $12/month lets you conduct longer symphonies. That price tag raised eyebrows, as Stable Diffusion is free and open source.
Salesforce's Einstein Copilot: The Chatbot That Seems Destined to Lose Over the Long Run
Salesforce gave their AI layer, Einstein, a makeover with Einstein Copilot. I wonder where they got that “Copilot” name from 🤔…
Ah, beats me.
Anyways, think of Salesforce’s Copilot as your assistant that's always ready to help, no matter which Salesforce product you're using. Clara Shih, the CEO of Salesforce AI, says it's like having a conversational AI sidekick to make your job easier and faster. Want to set up a storefront or handle a tricky return? Just ask Einstein Copilot in plain English.
Salesforce isn’t the only software company jumping on the generative AI bandwagon, and they believe their edge lies in covering a wider range of customer touchpoints.
That could be true over the short run, but the primary short term advantage Salesforce has compared to data warehouses like Snowflake or Databricks is speed to value. Users who are provisioned certain Salesforce data can immediately use Einstein Copilot, while you likely will not give the same users access to all of Snowflake. Provisioning all of that data for each user will take a long time.
Informatica’s chart highlights Salesforce’s longer term predicament, which I altered below for simplicity.
The purple arrow is Salesforce, which is just one application that an enterprise uses. All applications, and all other company data, is eventually fed into a data warehouse like Snowflake. Unsurprisingly, Snowflake already bought generative AI company Neeva.
Given that all company data, including applications like Salesforce, flows into Snowflake, data warehouses like Snowflake or Databricks seem positioned well ahead of Salesforce in the generative AI race.