In today's edition:

• Sam Altman downplays GPT-4

• Roundups of the top headlines and hottest product launches

• The ethical cost of training ChatGPT

• A 1923 cartoonist predicted generative art

🚂 GPT-4: The Hype Train

Everywhere you look, it seems like the AI community and Silicon Valley as a whole are having a bit of a meltdown over OpenAI's much-anticipated release of GPT-4. Rumors have been flying left, right, and center about its supposed capabilities, with some declaring GPT-4 as the next big leap in AI.

Quick refresher: GPT-4 is the successor to GPT-3 (duh) and ChatGPT, which is a fine-tuned model from the "GPT-3.5" series.

Before you get your hopes up, let's pump the brakes. In an interview with StrictlyVC, OpenAI CEO Sam Altman declared that "people are begging to be disappointed and they will be."

Furthermore, when asked about the viral (and false) chart comparing GPT-3's 175 billion parameters to GPT-4's 100 trillion, Altman simply called it "complete bullshit."

I've seen at least a dozen AI influencers post this exact chart. Altman must have a good reason to completely dismiss it instead of offering a better comparison between GPT-3 and GPT-4.

Still, let's not get too disappointed.

Altman also teased that a video-generating AI model is on the horizon, saying "It will come. I wouldn't want to make a confident prediction about when."

We may not be getting AGI or the Singularity anytime soon, but I'd love to see the first AI-generated cat videos.

AI technology is still in its infancy and OpenAI is working hard to prioritize developing its tools responsibly and ethically (more on that below). Let's all take a deep breath and try not to get too caught up in the hype.

Watch the full Sam Altman interview: (StrictlyVC)

🤠 Headline Roundup

Adobe said the company has “never, ever” used customer projects to train its generative AI models, in response to a wave of user criticism. (Bloomberg)

Google Research highlighted its progress in 2022 as well as its vision for 2023 and beyond. The advancements in Large Language Models, Computer Vision, Multimodal Models, Generative Video, and Generative Audio make it worth the read. (Google Research)

EletricNoir debuted Dark Mode, an AI-generated horror visual novel game. (VentureBeat)

Microsoft CEO Satya Nadella talked about AI revolutionizing work and balancing automation with wages for economic opportunity. (Video: The Wall Street Journal)

WIRED argued that AI is the next big thing for entrepreneurs. The internet gave us scalable distribution, while AI will give us scalable production, freeing up billions of hours of human labor and creativity. It's time to dust off that old business plan from high school. (WIRED)

Impel raised $104 million for its auto industry customer engagement service. (Crunchbase)

ChatGPT's Unsavory Underbelly

Apparently OpenAI picked the easy road to creating a "safer" AI chatbot. The company resorted to outsourcing Kenyan laborers earning less than $2 per hour to detoxify ChatGPT.

The decision was prompted by predecessor GPT-3's unfortunate tendency of spouting racist, sexist, and violent language. OpenAI's language models are trained on billions of words scraped from the internet, a vast repository of human language that unfortunately also includes a lot of bigotry and bias.

To combat this, OpenAI outsourced the labeling of tens of thousands of snippets of text to Kenya.

You can imagine the contents of the terrible texts scraped from the internet. Labeling them as violence, hate speech, sexual abuse, and so on was a necessary step to teaching the AI model to filter out toxicity.

This is a sobering reminder that the quest to build better AI tech comes not just at a financial cost, but an ethical one. 

In many cases, including this one, the cheapest solution can often have the biggest consequence on real people. These outsourced workers were not only paid less per hour than what we pay for a cup of coffee, but they were also exposed to graphic and disturbing content that will have a lasting impact on their mental health.

Whether this groundbreaking tech is being used for tweeting funny screenshots or producing impressive content, it comes with a heavy ethical price tag. The least we can do is learn and understand how exactly OpenAI built it.

(Time)

🔥 Hottest Product Launches

• Vowel - automated meeting summaries so nobody has to take notes (link)

• DoNotSign - reads the Terms and Conditions so you don't have to (link)

• Cursor - an IDE for developers optimized for integrated work with LLMs (link)

• Meet Millie - creates the perfect pick-up lines for your Tinder match's profile (link)

• Writer - a Figma plugin that creates copy so you can focus on design (link)

💼 Hiring Corner 

• Applied Researcher in Scene Understanding at Apple

• Machine Learning Engineer, Experimental AI Team at Instacart

• Machine Learning Researcher at Point72

• Machine Learning Scientist, Remote at UnitedHealth Group

• Research Scientist at Anthropic

Tweet of the Day

That's a wrap for today. Stay curious and see you tomorrow! If you want more bite-sized content, be sure to follow me on Twitter (@jeremykuoo).