Memory Leak — #37
VC Astasia Myers’ perspectives on AI, cloud infrastructure, developer tools, open source, and security. Sign up here.
🚀 Products
GPT-4o (“o” for “omni”) is a step towards more natural human-computer interaction. It accepts as input any combination of text, audio, image, and video and generates any combination of text, audio, and image outputs. It can respond to audio inputs in as little as 232 milliseconds, with an average of 320 milliseconds, which is similar to human response time (opens in a new window) in a conversation.
Why does this matter? AI voice technologies have lagged behind image and text. GPT-4o takes a profound step forward as a multi-modal service. The AI experience is far less monotonous and much more conversational than their predecessors in the voice-assistant world. We can imagine the next step will be additional emotional intelligence.
Introducing the First Ever Code Interpreter for R Based Analysis in Julius
Julius announced that users can analyze datasets, create charts and plots, perform computation, and apply statistical methods like regression, forecasting, PCA etc using AI and R code.
Why does this matter? Julius allows users to analyze their data with computational AI. It is a specialized service compared to ChatGPT, which is generalist. A day later OpenAI announced improvements to data analysis in ChatGPT including uploading files directly, interaction with tables and charts, and ability to download charts. This is a demonstration of the contrast between a product with a singular focus and one with a broader mandate.
Gemini Breaks New Ground With a Faster Model, Longer Context, AI Agents and More
At Google I/O there were a number of announcements including a series of updates across the Gemini family of models, including the new 1.5 Flash, our lightweight model for speed and efficiency, and Project Astra, our vision for the future of AI assistants.
Why does this matter? Personally, I thought Project Astra was the most interesting announcement because it can process multimodal information, understand the context you’re in, and respond naturally in conversation. This is a second nod to voice AI becoming more performant.
📰 Content
Should IBM Open Source Terraform?
In August 2023 Hashicorp announced that Terraform will be licensed under the Business Source License (BUSL). The important difference between BUSL and the previous Mozilla Public License (MPL) is that Hashicorp now retains the right to create commercial products from Terraform, unless a separate agreement is made. What this means in practice is that offering Terraform as a service, or as a significant part of a service that competes with Hashicorp’s own Terraform Cloud would no longer be allowed. This decision led to OpenTF (aka OpenTofu), a fork of Terraform that is open source, community driven, and managed by the Linux foundation. In April 2024, HashiCorp signed an agreement to be acquired by IBM for $6.4 billion. Given the change of control, people are wondering about the future of Terraform.
Last week, TNS VoxPop asked, “Should IBM open source Terraform?” Nearly 80% of respondents were in one of two “Yes” camps while 16% either don’t think IBM should give the code away or are happy with the OpenTofu fork.
Why does this matter? Hashicorp changing its license led to unease in the open source community. From the survey results it is clear users want the license to change again.
Bloomberg: Apple Finalizing Deal with OpenAI to Bring ChatGPT Features to iOS 18
Supposedly Apple is finalizing an agreement with OpenAI to bring some of its technology to the iPhone this year, according to a new report from Bloomberg. With this deal, the report explains that Apple will be able to offer “a popular chatbot” powered by ChatGPT as part of its AI-focused features in iOS 18. While Apple is also still in talks with Google about an AI partnership, tonight’s report says Apple has “closed in on an agreement with OpenAI.”
Why does this matter? While there is no guarantee that an agreement will be reached, it is interesting that these conversations may be happening. Apple was an early mover in voice AI with Siri. If Apple and OpenAI strike a deal that brings their new GPT-4o voice capabilities to the Apple AI suite it would be a demonstration that Siri has lagged behind.
Large Language Models (e.g., ChatGPT) as Research Assistants
Daniel Lemire, a computer science professor at the Data Science Laboratory of Université du Québec (TÉLUQ) in Montreal, writes about the rise of AI assistants in academia. Liang et al. found that an increasing number of research papers are written with tools like GPT-4 (up to 18% in some fields). He states, “It is quite certain that in the near future, a majority of all research papers will be written with the help of artificial intelligence. I suspect that they will be reviewed with artificial intelligence as well. We might soon face a closed loop where software writes papers while other software reviews it.”
Why does this matter? Research is a very popular use case for ChatGPT and other services. His reference to AI researching, writing, and reviewing papers reminds of a popular use case for multi-agent systems, an area we are very excited about.
💼 Jobs
⭐️DragonflyDB — React Tech Lead — Dragonfly Cloud
⭐️Chroma — Member of Technical Staff
⭐️Speakeasy — Product Engineer
Views expressed in posts and other content linked on this website or posted to social media and other platforms are my own and are not the views of Felicis Ventures Management Company, LLC. The posts do not constitute investment, legal, tax, or other advice and do not constitute an offer to invest in any security.
