Dall-E AI for Images; Direct Listings, SEC Ruling for Issuing Shares

We've got segments on artificial intelligence and IPO innovation in today's episode of 16 Minutes, where we take a look at the news and what it means for the long arc of innovation.

In the first segment (0:00): Take the surrealistic images of Salvador Dali and cross them with Pixar's animated film Wall-E and you've got ... Dall-E! It's a new neural network that creates images based on text inputs, and the worlds of A.I. and machine learning recently got their first glimpse.

Last summer, research lab OpenAI released an API for the machine learning model GPT-3, which caused a stir with the way it could produce text that was hard to distinguish from human writing (16 Minutes showrunner Sonal Choksi and a16z Operating Partner Frank Chen discussed it in a recent 16 Minutes Podcast, "GPT-3: Beyond the Hype," breaking down what it does and doesn't mean for startups, incumbents, and the idea of "AI as a service").

Now OpenAI has unveiled Dall-E, which processes language to create new images (not new text, as GPT-3 does). Dall-E does this using a neural network called CLIP (Contrastive Language-Image Pre-training), which classifies a wide variety of images culled from the internet while "filling in the blanks" using zero-shot reasoning, enabling Dall-E to produce surprising images by inferring information it wasn't trained in.

We called on Frank again to explain where Dall-E (and the broader topic of machine learning) sits on the path toward artificial general intelligence (AGI), how Dall-E's transformer-type architecture is able to infer information, what its limitations might be, and what uses we might see as this technology develops. - with Zoran Basich

In the second segment (12:58), we had a quick chat with a16z Operating Partner Scott Kupor about the recent decision by the SEC to allow the issuance of new shares via direct listings on the New York Stock Exchange. Previously direct listings were limited to the sale of existing shares. Recent first-day IPO "pops" have sparked much discussion about the fairness or unfairness of the process and whether the current path we have for companies going public is broken or just needs some tinkering around the edges.

Scott breaks down how the new rule will affect companies, as well as institutional and retail investors, and what this means in the long arc of IPO innovation. - with Zoran Basich