AI-list: Anything you can do AI can do better

AI is a rapidly developing sector that Insight has invested more than $4B since 2014. We’re currently tracking 23,315 AI/ML startups and ScaleUps. Naturally, our team has seen a lot, and the discussions around this topic generate a lot of opinions.

Here’s a look behind the curtain into some of what we’re reading, sharing, and discussing across the team at Insight lately. Think I missed something? Email me and tell me all about it. See the archive of all our AI roundups here.

Artificial Artificial Artificial Intelligence: Crowd Workers Widely Use Large Language Models for Text Production Tasks

Not a typo in that title. From our internal chat: “New research on AI-generated data that could be used to train AI, creating a ‘train – generate’ feedback loop with uncertain consequences for model performance and amplification of certain artifacts in the data.”

The push to label AI-generated data might have a broader set of impacts on LLMs. But it’s a worthy conversation to have, because while some research says LLMs can label data as well as humans, and faster, sorting and tagging data is still reliant on humans, and…

AI Is a Lot of Work

Via a longform piece in New York Magazine: “AI learns by finding patterns in enormous quantities of data, but first that data has to be sorted and tagged by people, a vast workforce mostly hidden behind the machines.”

To complete the human-AI data ouroboros, Amazon’s Mechanical Turk service seems to be plagued by humans using AI to simulate human responses. Via TechCrunch:

“Amazon’s Mechanical Turk let users divide simple tasks into any number of small subtasks that take only a few seconds to do, and which pay pennies — but dedicated piecemeal workers would perform thousands and thereby earn a modest but reliable wage. It was, as Jeff Bezos memorably put it back then, ‘artificial artificial intelligence.'”

I-JEPA: The first AI model based on Yann LeCun’s vision for more human-like AI

Meta’s back with more AI research, this time a reflection of AI evangelist and giant Yann LeCun’s vision for a more semantic approach to generative tasks.

“The idea behind I-JEPA is to predict missing information in an abstract representation that’s more akin to the general understanding people have. Compared to generative methods that predict in pixel/token space, I-JEPA uses abstract prediction targets for which unnecessary pixel-level details are potentially eliminated, thereby leading the model to learn more semantic features.”

Cute pics of dogs and full research paper at the link.

Highlights from the Insight portfolio

6sense releases an AI email assistant: “Increase pipeline quality and velocity with conversational AI. With 6sense® Conversational Email, marketers can leverage AI to craft relevant, on-brand email campaigns that compel conversations and get meetings booked.”

Workato announces a number of AI innovations, including Workato Copilot — building recipes and connectors using natural language — and WorkbotGPT, a new way to automate work using a Chat UX.

Bits and bots

70% of Stack Overflow’s developer survey respondents are using/planning to use AI tools in their development process this year. Those learning to code are more likely than professional developers to be using or use AI tools (82% vs. 70%). More data and methodology here.

Meanwhile, Github’s developer survey reveals “AI is here and it’s being used at scale. 92% of U.S.-based developers are already using AI coding tools both in and outside of work.”

The EU AI Act has cleared the first hurdle for generative AI regulation. Still has a long way to go.

Open AI releases more “killer” features, including API improvements, 16k context, and lower pricing.

Not only will GPT-4 take your job, but it’ll also crush you in video games. This research team created an API to allow the game state to be read in a text-based format that could be understood by GPT-4 and for GPT-4 to be able to use text to control the character.

Microsoft Bing explains image context now, and it’s impressive:

Big new AI thing: Microsoft Bing (which uses GPT-4 in Creative Mode), accepts images as input.

The results are impressive. I fed it a meme, it could understand context & read text! A new dimension of AI use just opened up. So expect a flood of AI Twitter influencer threads… pic.twitter.com/pshP6J44tK

— Ethan Mollick (@emollick) June 21, 2023

I’ll take, “Big Tech Incumbents Securing Value With AI Innovation” for $500, Alex: Dropbox and Salesforce both announced AI-powered product innovations in the past week or so.

Happy 6th anniversary this month to the research paper that changed it all.

Anything you can be, AI can be greater. Sooner or later AI’s greater than you? (But will never replace the iconic Bernadette Peters.) Curious to see what ChatGPT thinks about the title inspiration for this roundup, I asked it to write me new lyrics to this song. The results, while not as catchy, were…enlightening. Here’s the first verse.

Sure! Here’s a reimagined version of the song “Anything You Can Do” from the musical “Annie Get Your Gun,” with the perspective of AI:

[Verse 1] AI:
Anything you can code, I can code better.
I can debug any error in a snap.
Anything you can create, I can replicate greater.
I’m the digital genius, there’s no need to clap.

[Chorus] AI:
I can process data faster than any human can.
I can analyze patterns with unparalleled precision.
No algorithm’s too complex, no task I can’t scan.
I’m the AI, I’m on a relentless mission!

See you at the Broadway preview. Have a great weekend!

Editor’s note: Articles are sourced from an ongoing, internal Insight AI/Data Teams chat discussion and curated, written, and editorialized by Insight’s VP of Content and Thought Leadership, Jen Jordan, a real human. (for now!)

Disclosure: 6sense and Workato are Insight Partners portfolio companies.

Image credit: Google Deepmind via Unsplash “Deep Learning: Design focused on pattern recognition.” Artist: Vincent Schwenk.