A.I. Executive Briefing #3

The A.I. Executive Briefing is an expert weekly curation of A.I. news by our research team, shared externally now because we feel there’s too much hype & noise in the market. The same content will be distributed through?this substack.

News Round-up

1. Adobe is so confident its Firefly generative AI won’t breach copyright that it’ll cover your legal bills

2. ChatGPT in Mercedes-Benz, Apple Vision Pro, Google AI for Shopping

3. OpenAI releases function calling & Code Interpreter

4. Self-healing code

5. China's ByteDance Has Gobbled Up $1 Billion of NVIDIA GPUs for AI This Year

6. Voice AI is getting louder (bad pun..)

Venture News

7. Dropbox launches $50M AI-focused venture fund

8. Figma acquires?the AI design tool?Diagram

9. Voice-generating platform ElevenLabs raises $19M

10. Healthcare billing and claims automation startup Outbound AI raises $16M

11. Internal search engine startup AptEdge Closes $11M Seed Funding

News Round-up

1. Adobe is so confident its Firefly generative AI won’t breach copyright that it’ll cover your legal bills

Adobe Firefly, the software giant’s AI-powered image generation and expansion tool, is being rolled out to businesses this week. Adobe is so confident in?Firefly’s ability to respect creators’ copyrighted images that it announced that it will legally compensate businesses if they’re sued for copyright infringement over any images its tool creates. This is a bold claim, one in which I’m sure Adobe’s legal team did not appreciate but its marketing team did.

2. ChatGPT in Mercedes-Benz, Apple Vision Pro, Google AI for Shopping

苹果 announced its mixed-reality headset, Apple Vision Pro, on June 5. German automaker Mercedes-Benz AG announced that it will add ChatGPT to its cars via a beta program for Mercedes-Benz User Experience (MBUX) features in its vehicles. And 谷歌 announced new AI features for U.S. shoppers to virtually try on women’s tops from brands across Google, including Anthropologie, Everlane, H&M and LOFT. Each of these applications of A.I. — Apple vision pro (AR/VR), ChatGPT in Mercedez-Bens (Voice AI), Google Clothing AI (CV) — are nothing new. We’ve seen these applications tried many times in the past (e.g. Oculus, Amazon’s Siri in Ford cars, countless clothing AI companies that I was pitched 5 years ago). But is this time any different? Has the chasm between expectation + reality closed enough where the market is ready to adopt these applications or is this another hype cycle we’re going through?

3. OpenAI releases Function Calling & Code Interpreter

OpenAI released 2 potentially interesting tools over the past few days: Function Calling & Code Interpreter. Very simply, Function Calling now allows GPT-4 to execute functions when calling upon its API, which may pave the way to do things like check the weather, get a stock price, initiate another API, etc. If you’re a developer creating code at a company and using the GPT-4 API, I’m not sure what benefit this actually has at all, but it seems like this may be an important piece for the promise of autonomous agents. Code Interpreter in an alpha plugin that takes in a file, and uses GPT to write python to help analyze and answer questions about the file. A use case for this might be uploading a financial model and asking questions on that data.

4. Self Healing Code

The idea of using LLMs to write, edit and debug code is continuing to advance. Self-healing code is based on the idea that code can be treated as a living organism. This means that the code can learn and adapt to its environment. Although, self healing code is an advancement in the DevOps side of development (which is where GitHub 's developer survey demonstrated has the most efficiencies to be unlocked, see last weeks newsletter), this concept still requires human intervention for it to be truly effective. Unfortunately, for all the A.I. Agent maximalists, self healing code is a crutch for poorly written code to become passable by its given standards. Some Microsoft & MIT researchers proposed a hierarchy of GPT 3.5 for code production, GPT 4 for review and a human for a final assessment as the most ideal format for unlocking the most performance gains.

5. China's ByteDance Has Gobbled Up $1 Billion of Nvidia GPUs for AI This Year

In September of 2022, US officials instructed 英伟达 and AMD to stop selling their high-performance AI-focused GPUs to China and Russia without a license. However, 字节跳动 has already ordered around $1 billion worth of Nvidia GPUs in 2023 so far, which amounts to around 100,000 units split between Nvidia's A100 and?H800?cards, allegedly ordered prior to the ban that went into place toward the end of Summer last year. It's unclear if the $1 billion figure being reported refers to already-delivered or ordered-only shipments. This supply shock plus the staggering increase in demand has lead to an underground market of high performance GPUs & questions around who is supplying and where they are sourcing the technology from.

6. Voice AI is getting louder (bad pun..)

A.I. voice technology has been a highly watched and anticipated aspect of Generative A.I. Recently, we see IBM involving itself in this area by powering AI-generated audio commentary and captions in online highlights clips for the Wimbledon. We see this news on the back of the highly covered Eleven Labs fundraise (see below for more details) & off of real time voice modifications used by gamers powered by a company called VoiceMod. The biggest issue with generative voice technology, whether it be for real time translation or modification, is latency. As more voice data is utilized by these companies to build better models for their audiences, we will see latency incrementally tick down until near real-time deep fake voice tech truly may require new methods of verifiability such as cryptographically attested microphones.

Funding News

7. Dropbox launches $50M AI-focused venture fund

Dropbox enters the A.I. arena with a new universal search product that can search Dropbox files, Google Docs, Airtable, Slack, Salesforce and everything in between. In tandem they also announced a $50m A.I. fund to financially support the “Future of Work.” Other notable corporate VCs playing in the space are Salesforce Ventures with a $500 million generative AI fund, Workday?with a $250 million A.I. & ML expansion to its existing VC fund and?OpenAI with a $175 million fund to invest in AI startups.

8. Figma acquires?the AI design tool?Diagram

Figma acqui-hired Jordan, Siddarth, Andrew, Marco, and Vincent by acquiring Diagram, an A.I. powered Figma design tool founded in 2020 off the back of the GPT-3 release. A notable acquisition given Figma’s lack of A.I. development and the importance of Generative A.I., especially demonstrated by companies like Adobe with their Firefly product. A different breed when focused on generative product design or generative wireframes, especially when consider projects like Framer, which can spin up a website design, template, color scheme and sensical filler text for you right before your eyes only based on text input (think text-to-website).

9. Voice-generating platform ElevenLabs raises $19M

After 1 year, 15 person start-up ElevenLabs has achieved a $99m valuation to build out a hub for A.I. voice generation. The company’s AI text-to-speech models are language-agnostic, allowing corporate customers to fine-tune them and build their own, proprietary speech models on top. The company is Selling microphones and voices to those generating voices for gold (picks and shovel pun). As mentioned above, this space is noisy and will only get more cacophonous as companies clamor to clear the latency constraint. In addition to its current suite of voice architecting products, they have launched an API to allow companies to detect when a voice is generated from their platform itself. An intelligent (and potentially profitable) step in self regulating and promoting transparent A.I. development.

10. Healthcare billing and claims automation startup Outbound AI raises $16M

Outbound A.I. recently launched its?GPT-powered AI agent?to automate elements of the medical billing process, including updates to claim status, benefits verification, prior authorizations, and denials. Implementing A.I. to automate unsexy tasks with binary outputs will be where most of the efficiencies in the industry get actualized in the near term. Founder and CEO?Stead Burwell?said the new funding come as the US and other countries are bracing for a “dramatic increase” in patient volumes. He said factors such as the?greying of America, expanding middle class, and?nursing shortage?will contribute to the strain on healthcare workforces.

11. Internal search engine startup AptEdge Closes $11M Seed Funding

AptEdge allows users to connect their enterprise applications like slack & google drive to power an internal search engine. They are pioneering “AnswerGPT” to Revolutionize customer care. This is not the first nor the last company we have seen that is building internal search tools. I guess the question is here is how long until the open source movement enables businesses to build these tools themselves or will the non-technical remain taxed as they always have?

Reader Questions

Q: “Based on the Mistral AI funding news, are we really in an AI bubble? “

A: I think we’re starting to enter an AI bubble, but we’ll continue to see rounds + valuations like this over the coming months into 2024 IMO. I’ll share 2 snippets from a recent All-In podcast episode.

This was from Chamath Palihapitiya regarding Mistral's round:

This is not the way to make money, guys. I'm just going to be honest with you. So whoever's putting money and thinking they know what the fuck they're doing, you might as well just light it on fire, go to Vegas and have some fun with it because you will make more, you'll get more enjoyment from that than you will for making these kinds of investments.

But here's a more interesting insight from Brad Gerstner shortly after:

Let me say something different. In 1997, 98, we had a similar phenomenon. Everybody thought Internet search was going to be huge. There was massive FOMO and chasing, and everybody scrambled to get a search logo. Altavista, Infosys, Lycos, GoPlanet, Geo City. Just go through the list that people were scrambling after. And the truth of the matter is almost all those companies went to zero, even though you got a couple bets right. You got the Internet right, you got search right, but you didn't have to invest a dollar in search until 2003, and you would have captured 98, 99% of your time. Do it again, do it again now for social networking. Same thing.

VERY good point he brings up and the analogy to the internet search and social networking waves is spot-on. If we think AI is going to be as big as these past 2 waves, you can wait a few years, invest later, and still capture 90%+ of the value.

Q: “What do you think is the future of the ChatGPT plugin ecosystem? Will it be like Apple’s app ecosystem or will it die out?”

A: At the moment, the ChatGPT plugin experience is an absolute mess. I personally don’t think the ChatGPT plugin ecosystem will last — one of the reasons being that I think the business model doesn’t make sense. Companies such as Expedia are ultimately going to want their users to go to their site in order to collect that data and control that user experience.

Send us a message with any questions/comments/thoughts on anything A.I. related and we’ll try to answer them in our next release.