Lots of activity in AI land last week. Two themes: images and new players. Read on for more!

In the news this week:

• 🖼️ The next frontier in AI: images

• ➕ What’s up with Elon’s AI play?

• 🏥 An AI model specifically for healthcare

• ⚖️ The FTC digs into ChatGPT

📰 AI in the News

Meta and Google Tackle Images

Now that generating text is largely a solved problem, AI research labs are charging away at images. And both Meta and Google wanted to play last week.

Meta released CM3leon (pronounced “chameleon”), an AI model that can both generate/edit images as well as describe them. Normally these functions are split into different models.

Then, Google updated its ChatGPT competitor Bard to accept image uploads, opening up use cases like:

• Digitizing paper receipts

• Writing code to replicate a screenshot of a website

• Transcribing written notes on a whiteboard

We are still waiting for ChatGPT to release its own image capabilities. OpenAI had announced this back in March but has not given any detail about the public availability of these features.

Still, it’s clear that teaching an AI to “see” is the next hot area of focus. What use cases are on the horizon?

Claude 2 and Elon Musk’s xAI

One thing to expect for the 12+ months: the AI landscape is going to get even more crowded.

Let’s start with “second-place” AI model startup Anthropic. Their AI models are second-best behind ChatGPT, built by a team founded by ex-OpenAI employees.

Anthropic just released Claude 2 to the public. It’s the first time the general public is able to play with their chatbot, after an extended beta for Claude 1.

Claude 2 is notably more reserved than ChatGPT, unwilling to do some tasks like rewriting copy in the name of safety and harmlessness. But, you can upload PDFs that are very long (e.g., 75 pages!) to Claude 2, something you can’t do with ChatGPT.

Across ChatGPT, Claude, Bing Chat, Google Bard and many more coming, the average user is going to face a challenge in remembering which chatbots are good for what purposes.

Add on the mysterious xAI, Elon Musk’s new AI lab venture announced last week. There’s very little known about what xAI will focus on. One guess: math.

Cryptic tweets from xAI’s inaugural team members suggest that building an AI that can master the field that ChatGPT is notoriously bad at could result in a fundamental unlock for AI.

It’s an ambitious goal - but that’s what you should expect from Elon Musk these days, right?

More Stories

• KPMG is partnering with Microsoft to invest $2 billion in AI efforts. (link)

• IT outsourcing giant Wipro is investing $1 billion to train its staff in AI. (link)

• Israel quietly puts AI systems at the center of military operations. (link)

• OpenAI is paying the Associated Press and Shutterstock to train models on their proprietary data. (link, link)

• The FTC has opened an investigation into whether or not ChatGPT causes consumer harm by generating false information. (link)

• Shopify is giving its sellers an AI assistant called Sidekick. (link)

💡 Industry Insights

🏥 Healthcare

AI x nursing: The University of Minnesota is forming a group to explore the use of AI in nursing. The backdrop: the nationwide nursing shortage. On one side, optimism that AI could help lighten the administrative load on nurses. On the other side, skepticism that current AI capabilities would help more than hurt. (link)

Hippocratic AI: To address nurses’ concerns, startups like Hippocratic AI think that purpose-built AI models will be the key. Hippocratic’s model is narrowly tailored for healthcare, trained on high-quality and legally acquired healthcare data. (link)

AI in pharma sales: Pharmaco investment in AI is expected to reach a whopping $3 billion in 2025. Marketing and sales executives see AI as a key to deeply personalized commercialization and marketing processes. The goal: targeting more relevant people with more relevant information. (link)

👩‍⚖️ Legal

AI vs. publishers cont.: You gotta pay us, they say. Social media platforms like Twitter and Reddit now charge millions for access to their content. OpenAI has had to strike paid partnerships with the AP and Shutterstock to use their data. Now, media holding company IAC plans to sue AI model makers. (link)

FTC investigation: The FTC wants OpenAI to explain how it collects and uses user data. And, it wants OpenAI to explain how it’s dealing with hallucinations (AI models making up information). Zoom out: AI moved fast for about 6 months, and now consumers/broader society are reacting just as quickly. (link)

📈 Venture Deals

• Prolific, which stress-tests AI models, raised $32 million. (link)

• Simbe raises $28 million Series B for robotics and AI in retail operations. (link)

• Slang, an AI phone-answering service, raised $20 million. (link)

• Open source AI model creator Nomic AI raises a $17 million seed round. (link)

• Immersive Fox, a software for producing videos of your face, raised $3.3 million. (link)

🛠️ Latest AI Tools

• Coframe is an AI assistant that operates and self-improves your website 24/7. (link)

• DragGan lets you completely manipulate photos via point-and-drag controls. (link)

• AskAI can answer questions about your content/docs in Slack and Teams. (link)

• Floneum is a graph editor designed for AI workflows. (link)

• Spoke AI summarizes busy Slack channels from within the app. (link)

• AI Companion App is a template to create and chat with AI companions. (link)

• Gloo searches and cites your data via semantic search. (link)

Share & Subscribe

Did someone send you this newsletter? Subscribe Here.

Know someone who’s interested in learning about AI? Copy and paste this link to others: https://tiferes.beehiiv.com/subscribe