Daily Digest: Trouble at OpenAI

PLUS: Gemini 1.5 Flash is getting popular.

May 20, 2024

Want to get in front of 100k AI enthusiasts? Work with us here

Hello folks, here’s what we have today;

PICKS

  1. OpenAI had a troublesome weekend, again. In his resignation, Jan Leike spoke about the company’s reduced focus on AI safety, while uncommon equity practices are also inviting criticism. Sam Altman (CEO) and Greg Brockman (President) addressed both concerns over the weekend. 🍿Our Summary (also below)

  2. OpenAI has also paused the voice of Sky in ChatGPT’s voice mode—most likely due to its high resemblance to Scarlett Johansson’s voice. OpenAI claims the voice belongs to a voice actor and outlines how the voices in ChatGPT were selected.

  3. Gemini 1.5 has an updated technical report. The report has good technical details but here’s the crux:

    • Gemini 1.5 Pro is similar to Claude 3 Opus but inferior to GPT- 4o.

    • Gemini 1.5 Flash performs similarly to Claude 3 Sonnet, but faster & cheaper than Haiku.

    • Google has also built two specialized models: Gemini 1.5 Pro for math, which scores 91.9% on the MATH benchmark (insanely high) and Flash 8B which is even smaller than Gemini 1.5 Flash.

from our sponsor

Want to Add Hospitality AI to Your Portfolio?

For a limited time, Ben's Bites readers can invest alongside top VCs and 1,200+ individuals in the AI transforming the $4.1T hospitality industry.

Partnered with Airbnb, Vrbo & Expedia, Jurny automates operations for thousands of property managers across the globe.

5x customer growth, $35M+ bookings last year

Featured in Skift, Forbes & Bloomberg

TOP TOOLS
NEWS
QUICK BITES

There's some drama at OpenAI, again. Safety researchers are leaving and questioning the company, while uncommon equity practices are inviting criticism. Moreover, it is pausing an AI voice in its products soon after demoing a real-time voice assistant.

What is going on here?

OpenAI is in the tangles of controversy again regarding the safety of AI models.

What does this mean?

It all kicked off when Ilya Sutskever, their former chief scientist, and Jan Leike, the head of the superalignment team, left the company. A few other team members followed suit.

Jan spoke out about his resignation, saying it felt like an uphill battle doing safety research at OpenAI. He was also unhappy with the company's shift towards building "shiny products" instead of focusing on their mission—making sure AGI is safe.

Meanwhile, Kelsey Piper @Vox released a piece on OpenAI’s offboarding agreement. It prohibits ex-employees from talking foul about OpenAI. If they do, they risk losing all their vested equity.

Greg Brockman, the president of OpenAI, responded to Jan's comments by explaining that the company's current safety strategy involves gradually releasing their products and systems into the real world to gather feedback and make improvements.

As per a Wired report, what’s left of the Superalignment team is being absorbed into other groups at OpenAI with co-founder John Schulman overseeing the far-fetched risks of AGI.

Sam Altman, the CEO, also addressed concerns about a clause in their exit agreements that could potentially take away employees' shares. He admitted that the clause was a mistake, stating it had never been used and inviting any former employees who might be affected to contact him directly.

Why should I care?

Well, OpenAI presents itself as the go-to place for building safe artificial general intelligence (AGI). These incidents damage that image—not just among regular users, but also among researchers and those who regulate AI technology.

Although, the vibes feel like “OpenAI has lost the mandate of heaven” such drama is not unexpected. OpenAI is the hottest company in the race to AGI, so it’s naturally going to attract attention—both positive and negative.

As this drama dies down, OpenAI is now facing another challenge. They've paused the use of Sky’s voice in ChatGPT, likely because it sounds too similar to Scarlett Johansson's voice.

Ben’s Bites Insights

We have 2 databases that are updated daily which you can access by sharing Ben’s Bites using the link below;

  • All 10k+ links we’ve covered, easily filterable (1 referral)

  • 6k+ AI company funding rounds from Jan 2022, including investors, amounts, stage etc (3 referrals)

Reply

or to participate.