• FF Daily
  • Posts
  • Forward Future News Episode 10

Forward Future News Episode 10

Nvidia, OpenAI, New Open Source Models, and more

This was an exciting week for AI. Nvidia, which makes GPUs that have historically powered video games, is dominating the news for many reasons. Let’s see what they’re up to!

Image: NVIDIA ACE for Games Demo

Nvidia Joins The $1T Club 📈

Nvidia this week joined the trillion-dollar company club because of its considerable investments in AI. Jensen Huang, the CEO, and co-founder of Nvidia, has made it clear that they are betting the company on AI. The good thing for them is AI models take a LOT of math to train and run, and GPUs work incredibly well for this use case.

Now, Nvidia joins an elite group of companies that have reached a $1,000,000,000,000 market cap, which includes Google, Microsoft, Amazon, and Apple.

Nvidia is not a new company. They have been around for decades. However, because of a massive influx of orders from companies wanting to train their models, they recently forecasted a ~25% increase in revenue, which is insane for a company of their size.

Nvidia Demos AI-Powered Video Game 🎮

During Jensen Huang’s first keynote in years, he demoed something I’ve been excited about for a long time. Using Unreal Engine plus Convai technology, he showed off an entire conversation between an NPC and a player. What was unique about that conversation was the NPC was entirely powered by AI, meaning it could have a dynamic conversation about pretty much anything within the context of the game.

AI and video games are coming together quickly. I’ve made two videos about it recently. I’m also planning to do a deeper dive into Convai, as I’ve been speaking with the founder and have been testing out his technology.

Here’s the demo video:

Lawyer Uses ChatGPT and Cites Fake Case 😂

A lawyer is in big trouble after getting caught using ChatGPT for his case, which cited six fake cases. This is a big reminder that ChatGPT is far from flawless and often hallucinates.

Lawyer Steven Schwartz of the firm Levidow, Levidow, & Oberman "greatly regrets having utilized generative artificial intelligence to supplement the legal research performed herein and will never do so in the future without absolute verification of its authenticity," Schwartz wrote in an affidavit on May 24.

This reminds me of over-relying on Wikipedia, which can be wrong because it is edited by…anyone. ChatGPT can also get things wrong so always take facts from ChatGPT with a grain of salt.

OpenAI Discusses Its Plans

OpenAI is currently limited on GPUs, and it’s causing a lot of problems for their short-term plans. This GPU bottleneck has delayed some of their plans, including:

  1. Improving their API speed

  2. The wide rollout of 32k context windows

  3. Dedicated capacity for specific paying customers

What is on their short-term roadmap?

  1. Cheaper and faster GPT4 - they aren’t working on GPT5 yet, so they are hyper-focused on making GPT4 the best product possible.

  2. Increased context windows - a common request from users wanting longer prompts.

  3. Adding memory to the API - no longer having to pass context on every API call.

  4. Fine-tuning API endpoint.

Sam Altman mentioned plugins don’t have product market fit. He says they are cool but doesn’t necessarily use them himself. He believes more people want ChatGPT inside of apps rather than apps inside of ChatGPT.

The OpenAI team also published a blog post titled Governance of Superintelligence about their position on regulating AI. In the blog post, they called for 3 main pillars of regulation: 1) coordination between AI leaders 2) formation of an agency like the IAEA (international atomic energy agency) for AI, and 3) continue working on the technical capabilities to “align” AI with humans.

They also mentioned their desire to keep open-source projects free from regulation but at a certain level of capabilities, they probably will want some red tape.

Quick News

  • Japan Goes All In on AI - the Japanese government affirmed that it will not enforce copyright on data used in training AI.

  • Majority of Americans Have Heard of ChatGPT - A study from Pew Research showed that a majority of Americans have heard of ChatGPT, but few have actually used it.

  • UAE Launches Falcon 40b Model - It immediately took the #1 spot on the LLM leaderboards.

How to reach me:

New Videos From This Week:

Reply

or to participate.