Welcome to AI Future - your weekly one-stop shop for all things AI.

Big shakeup in AI from Meta last week. And they did it by giving something away for free. That’s right.

In the news this week:

• 🦙 Meta’s LLaMA 2

• 👁️ ChatGPT’s woes

• 🛡️ Biden administration’s eyes safety measures

• 😷 AI in hospitals and curriculums

📰 AI in the News

Meta’s LLaMA Takes On ChatGPT

The AI ecosystem is witnessing a vital standoff between open and closed-source models.

This week Meta unveiled LLaMA 2, its fully open-source answer to ChatGPT. LLaMA 2 offers a robust, free alternative comparable to GPT-3.5 and open for tweaks. This means businesses can use LLaMA 2 as a “base” model and add their data on top of it.

In the other corner, OpenAI is feeling the heat. There's growing concern that ChatGPT's capabilities have slipped.

A Stanford study confirmed this theory, indicating a drop in math, reasoning, and coding abilities on some tasks since March. The exact cause of this drop remains unclear, but OpenAI has vowed to investigate.

Businesses who want to customize their own models don’t want their vendors suddenly making their AI models less capable overnight. For those businesses, building on top of LLaMA 2 makes a ton more sense.

Top AI Companies Pledge Safety

The Biden administration is finally putting AI safety in the spotlight. Ensuring AI builders sing from the same hymn sheet. Just last week, seven titans of AI—OpenAI, Anthropic, Google, Inflection, Microsoft, Meta, and Amazon—vowed to adhere to 'voluntary' safety guidelines, including:

1. Conducting security tests before AI system releases.

2. Sharing AI risk information.

3. Investing in cybersecurity.

4. Encouraging third-party vulnerability reports.

5. Identifying AI-generated content.

6. Openly discussing AI societal risks.

Notably missing from the commitments is any sort of disciplinary action if the listed companies don’t follow the rules.

Another subject missing from the conversation: Apple. The tech giant appears to be taking its sweet time to infuse AI into its flagship products (or so it seems).

This week, they announced an internal chatbot in the works — dubbed by some as “Apple GPT” — but has no plans to integrate it into any consumer products anytime soon.

More Stories

• Google is pitching media companies for an AI that can write news articles (link).

• Google DeepMind is using AI to develop more efficient AI computer chips (link).

• The consulting giant SAP has invested in AI startups Anthropic, Cohere, and Aleph Alpha (link).

• Qualtrics plans to spend $500 million on AI over the next four years (link).

• More than 8,500 authors signed an open letter to be compensated for their work (link).

• The UN Security Council will hold its first AI safety discussion in New York (link).

• OpenAI is worried that ChatGPT with visual search could be used for facial recognition (link).

💡 Industry Insights

🏥 Healthcare

AI x telehealth: Teladoc Health is partnering with Microsoft-owned Nuance to integrate voice-enabled generative AI into their system, aimed at assisting with documentation. The clinician-patient interactions will be converted into clinical notes using tools powered by GPT-4. Initial access is planned for select doctors and clinicians by year-end (link).

Chatbot consultation: K Health, a seven-year-old startup, is deploying an AI chatbot that absorbs patients' symptoms and medical history, compares them with millions of other patient records, and suggests a possible condition. The chatbot has facilitated a whooping 3.1 million chats so far (link).

AI for med students: Beth Israel Deaconess Medical Center in Boston is using GPT-4 to train medical students on critical thinking in diagnosis. Students are encouraged to treat the AI like a professional consultation but also be wary of over-reliance or misinterpretation (link).

👩‍⚖️ Legal

AI vs. artists: A federal judge is expected to throw away most claims in a lawsuit accusing Stability AI, Midjourney, and DeviantArt of training their models on artists’ work (without permission). The judge finds it “implausible” that the plaintiffs were actually infringed upon (link).

Detecting NYC fare evaders: The MTA in New York City is instituting an AI surveillance system to catch people who avoid paying fares. The third-party tracking software has been used at seven locations so far and will slowly roll out to two dozen more by the end of the year (link).

📈 Venture Deals

• Causaly, an AI platform for discovering and testing drugs, raised $60 million (link).

• AI and metaverse platform Futureverse raised $54 million (link).

• Cognaize, a platform for financial AI applications, raised $18 million (link).

• Nomic, which helps devs deploy leading AI models locally, raised $17 million (link).

• Humanoid robot builder Figure raised another $9 million from Intel Capital (link).

🛠️ Latest AI Tools

• Comigo is a cognitive AI assistant for people with ADHD (link).

• Tome generates slide decks from text prompts (link).

• Chathub lets you simultaneously use ChatGPT, Bing, Bard, Claude & more (link).

• Rightpage’s Zendesk app provides agents with quality responses to support tickets (link).

• Gradio is a platform to build and share machine learning apps (link).

• Leap helps you add AI to your product with a single API (link).

• Bifrost scans app designs and converts them into code (link).

• Broadcast automates your team’s weekly updates via Slack/email (link).

• FYI is will.i.am’s AI copilot for creatives (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