Jen Jordan
More By The Author
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.
Google Tests A.I. Tool That Is Able to Write News Articles
Per the NYT, Google is pitching news organizations, including the New York Times, Washington Post, and News Corp, on a tool called “Genesis” that can generate news articles by taking in details of current events. Presented as a journalist aid versus replacement, the obvious benefits and caveats apply. For understaffed newsrooms, this could be a solution for tight budgets. But so far, AI lacks the ability to understand cultural nuance, and there many instances of hallucination and issues with fact-checking.
Related: This week, the American Journalism Project announced a new partnership with OpenAI to support local news with their technology and a $5M grant. This comes on the heels of an agreement for OpenAI to license some of the AP’s archive dating back to 1985 to help train its algorithms.
WormGPT: New AI Tool Allows Cybercriminals to Launch Sophisticated Cyber Attacks
We’re officially in the blackhat GPT era. Per Hacker News: “In the hands of a bad actor, tools like WormGPT could be a powerful weapon, especially as OpenAI ChatGPT and Google Bard are increasingly taking steps to combat the abuse of large language models (LLMs) to fabricate convincing phishing emails and generate malicious code.”
What AI can do with a toolbox… Getting started with Code Interpreter
Since I last wrote this roundup, so much has come out about how truly amazing Code Interpreter is. This write-up via Ethan Mollick has a ton of great examples, and this gem: “Code Interpreter continues OpenAI’s long tradition of giving terrible names to things, because it might be most useful for those who do not code at all. It essentially allows the most advanced AI available, GPT-4, to upload and download information, and to write and execute programs for you in a persistent workspace. ”
(No) one foundation model to rule them all
From Matthew Lynley, “Claude and GPT-4 might seem like competing interpretations of how LLMs should work. More companies are using both, for different use cases.”
How it works in real life, per Insight portfolio company Rasgo CTO and co-founder, Patrick Dougherty: “Claude powers a couple things in our product as well, it either did better than GPT-4, or it worked faster, or we needed the 100,000-token context window,” Dougherty said. “It was really interesting, because the reasoning that it does through the AutoGPT style approach, it was pretty good—but the SQL generation was not as good.”
State of AI Development: 34000% growth in AI projects, OpenAI’s dominance, the rise of open-source, and more
Via Replit’s blog, some amazing AI builder stats: “Since Q4 of 2022, we have seen an explosion in AI projects. At the end of Q2 ‘23, there were almost 300,000 distinct projects that were AI related.”
Bits and bots
Llama-v2 is out, open source, with a license that authorizes commercial use.
This is huge: Llama-v2 is open source, with a license that authorizes commercial use!
This is going to change the landscape of the LLM market.
Llama-v2 is available on Microsoft Azure and will be available on AWS, Hugging Face and other providersPretrained and fine-tuned…
— Yann LeCun (@ylecun) July 18, 2023
Langsmith is out: “Today, we’re introducing LangSmith, a platform to help developers close the gap between prototype and production. It’s designed for building and iterating on products that can harness the power–and wrangle the complexity–of LLMs.”
New research shows another similarity between AI and humans: we don’t like to read. “We find that performance is often highest when relevant information occurs at the beginning or end of the input context, and significantly degrades when models must access relevant information in the middle of long contexts.”
Elon Musk has launched a new AI company called xAI, which I genuinely thought was the name of one of his children.
AppleGPT? Not to be left out, Apple is (finally?) reportedly testing an LLM and chatbot called “Ajax.”
Clippy is back, baby. But it’s not through Microsoft’s doing. Just glad someone shares my Clippy enthusiasm.
Happy Barbenheimer weekend to all who celebrate. A little uncanny valley, but entertaining nonetheless, from AI filmmakers Curious Refuge.
ScaleUp:AI announces speakers
Insight’s leading AI event, featuring top AI thinkers, leaders, investors, and innovators, is back this October. We just announced the first few speakers, including GitHub COO Kyle Daigle, Goldman Sachs CIO Marco Argenti, two co-authors of the seminal AI paper, “Attention Is All You Need,” and many more to be announced over the coming weeks.
AI-list readers can get *free* virtual tickets by using the code AILIST and registering here.
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.
Image credit: Google Deepmind via Unsplash.
