Sunday Links: Data workflow, Video Magic, and AI Autumn Leaves

Here are this week's AI links:

• Netflix Open Sources Maestro, a Next-Gen Data Workflow Engine. Another example of big industry players releasing powerful open-source. Maestro is a highly scalable data orchestration layer for AI/ML systems. This isn't as headline-grabbing as an LLM, but still very valuable. Netflix has a history of releasing interesting open source that works at scale - see their collection of open source here.
• Introducing SAM 2: The next generation of Meta Segment Anything Model for videos and images. Continuing the theme of open source, Meta added to its stack with an extremely slick-looking video segmentation model. The system enables you to use prompts to separate out and independently track items in a video. It also works for other things, but the video capability is the headline grabber here. This is extremely hard to do and opens up huge opportunities in video editing and processing. Expect your Instagram Ad feed to get increasingly wild in the next couple of years as these capabilities make it into tools.
• “Death of a Salesforce”: Why AI Will Transform the Next Generation of Sales Tech. Fresh off the clickbait headline printing press, A16Z's team gives a well laid out structure for how AI may transform sales. Some things I definitely agree with (blending sales, marketing, and customer success [also support more generally], but for me, this doesn’t go anywhere far enough. It doesn't actually suggest that Salesforce is going away; rather, it suggests that tooling layers will change. The big miss, though, is that it doesn’t talk about the buyer side evolution at all, which is where I expect massively more change. All the channels (SEO, Ads, content marketing, etc.) are going to be massively affected by AI. Buyers I think will have their own AI agents to get and maintain vendor relationships. That future will look extremely different.
• Google’s hiring of Character.AI’s founders is the latest sign that part of the AI startup world is starting to implode. The team's post is online here. The Fortune article adds a bit of AI gloom, but it's not surprising that there are shakeouts coming in sectors of the AI space that have seen 10+ funded competitors. Character.AI also had a very broad vision, which has turned out to be hard to monetize. It was unclear how they would maintain a technical lead with so much innovation happening. The B2C-focused market approach hasn't gotten enough traction to monetize, and B2B cases (e.g., characters in games) likely need very specific solutions. Elsewhere, Meta has scrapped plans to make further celebrity chatbots, which have fallen rather flat [Paywalled].
• State of AI in Venture Capital 2024. Michael Spencer and Alex Irina Sandu at AI Supremacy have a timely and well-organized list of where the biggest funding rounds in AI have gone. Their take (surprise, surprise) is also that 2024 may turn out to be the peak funding year for Generative AI. I'd say they are almost certainly wrong about that. Not because there won't be a dip (which seems highly likely) but because Generative AI (and AI more generally) seems like it will obviously eat a large part of future software development. Future years will see this initial wave spread to almost all parts of VC investment in software.

No doubt there will be a lot of hand-wringing that Generative AI may see as a "winter," but like the hype wave before it, it will be exaggerated. The technology has clear value; what we'll see now is who wins (probably due to their business leverage and execution rather than the technology).

If you're in Barcelona in September, Nesy 2024 will be taking place there. The future is Neurosymbolic, so I would like to check out if you can!