Welcome to the AI Index Report, featuring practitioner analyst and entrepreneur Toni Witt. This series is intended to deliver the timely intelligence about artificial intelligence (AI) you need to get up to speed for an upcoming client engagement or board meeting.
This episode is sponsored by the AI Summit Preconference at Community Summit, taking place October 16th in Charlotte, NC. The full-day preconference is dedicated to providing an understanding of where, why, and how AI should be applied to drive business results.
Especially among enterprise customers, open models have been gaining momentum. French AI startup Mistral AI is springing up to meet this need. It recently released its first model called Mistral 7B — 7B meaning 7 billion parameters, which is a relatively small, lightweight model size.
This model is available for download in various means. The startup also started a GitHub repository and a discord channel and put the model on Hugging Face. The model is under the Apache 2.0 license, which lets people use the model for free as long as they attribute the creators and have the compute resources. This current setup wouldn’t really work for large models.
While Mistral AI doesn’t charge for people to use the model, it does offer a paid version providing access to the model weights and training data. Open models have the advantage of giving more customization and flexibility, limiting security risk, and avoiding the black box problem or other ethical concerns.
Amazon recently announced that it’s going to invest up to $4 billion in cash in the AI startup Anthropic in an effort to compete with the other cloud rivals on AI investments. This deal includes priority and early access to its technology. Anthropic committed to relying on AWS and buying chips directly from Amazon for future training needs.
The biggest reason behind this investment is that Amazon wants to retain customers of their cloud services by keeping its AI offerings in-house and competitive. The startup said the further investment is not compromising its mission of building AI models that actually adhere to moral values and are generally good for humanity.
Product of the Week (07:55)
The great debate about which image generators are the best continues, with Midjourney and Stable Diffusion being widely considered superior to OpenAI’s DALL-E series. However, OpenAI has been stronger on the technology side. Now, it released DALL-E 3 which might change the direction of this debate.
OpenAI recently previewed DALL-E 3. It’s accessible for ChatGPT Plus and enterprise customers. With this new version, OpenAI focused on improving the mapping from prompt to output. OpenAI wants its offerings to be accessible to everyone and that’s the direction it’s continuing to develop the DALL-E series.
The company is building a product ecosystem rather than separate products, which is an element that differentiates it from its competitors. DALL-E can render texts, and other image generators typically struggle with that. The direction OpenAI is heading is building this product ecosystem that’s naturally multimodal — building generative AI models that can mix text with image with sound with video and more.