Howdy, wizards.
There’s been a storm of seemingly justifiable criticism towards OpenAI over the weekend. There’s the poor treatment of Scarlett Johansson, ex-superalignment lead Jan Leike speaking out about safety concerns, and some questionable equity practices coming to light.
Here’s the drama brewing in AI this week:
- ChatGPT gets improved data analysis skills. Businesses with data you’re not sure what to do with – here’s your new favourite feature.
- OpenAI takes down the “Sky” voice following scrutiny from Scarlett Johansson’s attorneys. And it’s a bad look for Sam Altman.
- OpenAI dissolves its safety team. Ex superalignment lead Jan Leike shines light on why he left.
- Controversial clause in OpenAI’s off-boarding agreement. A mistake, according Sam Altman.
- Microsoft’s “AI-first” PC. Faster AI, which remembers everything and is available throughout your apps.
Dario’s Picks
1. ChatGPT’s improved data analysis features
Let’s start this newsletter on a positive note before diving into the latest scandals.
OpenAI is rolling out a some highly useful features for data analysis to paid subscribers over the coming weeks:
- Upload files directly from Google Drive and Microsoft OneDrive.
- An expandable view that lets you do work on tables and charts in real-time. You can add a dataset, which ChatGPT turns into an interactive table or chart which updates as you analyse the data.
- The ability to customise charts to use in docs and presentations.
Why it matters
For those of you who have already discovered the power of using ChatGPT for data analysis, this is a real treat. I think it’s also likely to be a hit with companies, it can help making the process of getting insights from their data more efficient, and also make data analysis accessible to more people using plain English.
There’s also big potential of easily automating a lot of reporting work using the new analysis features + custom GPTs.
2. OpenAI pulls the “Sky” voice from ChatGPT
OpenAI demoed its new voice mode last week but drew attention to more than just how good it sounds. Many had already noted that when ChatGPT’s voice mode was first released in September, the voice “Sky” was eerily similar to Scarlet Johansson’s voice.
Now, a statement by ScarJo reveals that Sam Altman contacted her a year ago to lend her voice to ChatGPT, but she declined.
He told me that he felt that by my voicing the system, I could bridge the gap between tech companies and creatives and help consumers to feel comfortable with the seismic shift concerning humans and AI. He said he felt my voice would be comforting to people.
Scarlett Johansson, statement shared by @bobbyallyn
Two days before last week’s demo of the new voice mode, Sam Altman again contacted Johansson’s agent to reconsider. But before they managed to connect, OpenAI went ahead and released the demo and the voice sounds even more similar with added warmth and emotion.
Scarlett Johansson hired attorneys who questioned OpenAI about the process in which their Sky voices had been created, after which OpenAI took down the Sky voice. In their announcement, they also say the voice belongs to a different voice actor.
Context: Sam Altman’s favourite movie is apparently Spike Jonze’s 2013 sci-fi Her, in which Scarlett Johansson plays the voice of an AI assistant that the protagonist falls in love with. Right after the demo of the new voice mode last week Altman’s post on X read simply “her”.
Why it matters
It’s baffling how poorly this situation seems to have been handled by Sam Altman; clearly not good for his image. I saw someone ask this forward-thinking question: if a rich & well-connected person like Scarlett Johansson gets this kind of treatment from OpenAI, then what can we expect for the individual rights for the rest of us?
3. OpenAI dissolves its safety team
The “superalignment” team at OpenAI has been dissolved following the departure of several researchers in last months, most recently chief scientist and co-founder Ilya Sutskever and superalignment lead Jan Leike. This was the team primarily responsible for researching how to keep AI under control as it gets more advanced.
Leike initially announced his resignation with a short message last week: “I resigned”. However, he came back to X on Friday sharing a thread with details about his decision. He mentions disagreement with OpenAI’s leadership over core priorities, and not enough compute resources allocated for his team do carry out their job.
Why it matters
Recent events and statements from ex employees make it quite clear that the most safety concerned individuals working at OpenAI are leaving the company – they no longer trust that OpenAI is taking the necessary precautions in developing AGI.
The implosion of the safety team puts OpenAI, and Sam Altman in particular, under more scrutiny and distrust from just about everyone who cares about how the future with AGI might look like.
4. OpenAI’s controversial off-boarding agreement
Vox published an article this weekend detailing both the internal safety-concerns and a highly restrictive off-boarding agreement that OpenAI employees are made to sign. Essentially, the contract forbids them, for the rest of their lives, to criticise their former employer or even acknowledge that the agreement resist. If they refuse to sign it, they may lose the equity in the company they’ve earned up while working there – oftentimes millions of dollars. In other words, a very effective measure to keep everyones mouth shut.
Sam Altman responded to this criticism on X, saying essentially that it was a mistake, that they’ve never clawed back anyone’s equity and that they were in the process of changing/removing thee clause regarding potential equity cancellation.
Why it matters
I see some people are cutting OpenAI slack on this one blaming it on lawyers, and that CEOs don’t have time to read all the small letters. That could be true – but isn’t it strange that nobody in the leadership team caught and changed this until it ended up as public knowledge?
5. Microsoft’s “AI-first” PC
Microsoft just introduced something they call Copilot+ PCs, starting at 999$. Here’s what’s different about them compared to a regular PC:
- 20x more powerful and up to 100x faster for running AI workloads
- It has Windows 11, where Copilot assistant is now widely integrated with the OS itself
- It has a new feature called Recall allowing users to search for anything they’ve previously seen on their screens (reminds me of what Rewind is doing).
- Copilot will have screen-sharing that allows AI see what you’re doing, and chat with you about it (basically the same as the new ChatGPT desktop app for Mac)
Why it matters
It’s a good sneak peak into how AI will probably be integrated across all our devices very soon: speedy, across apps, deeply personal and it can see what you’re doing.
'; $(this).html(flexWrapper); }); } else if (index === goodBadColumnIndex || index === pricingColumnIndex) { $(this).addClass('desktop-only'); } }); }); }); $htmlContent.find('table').wrap('
'); $('.blog-post-rich-text').html($htmlContent.html()); $('.blog-post-rich-text a').attr('target', '_blank');});
$(function() { const path = window.location.pathname; const pageTitle = path.split('/').pop(); $('table .tool-name-wrapper').each(function(index) { const anchorId = `#anchor-${index + 1}`; const fullUrl = `https://www.whatplugin.ai//blog/${pageTitle}${anchorId}`; const jumpLink = $('', { href: '#', text: 'Jump to review', class: 'jump-to-review', click: function(e) { e.preventDefault(); const isMobile = window.innerWidth <= 768; const mobileOffset = 50; // Adjust this value for mobile const desktopOffset = 80; // Adjust this value for desktop const offset = isMobile ? mobileOffset : desktopOffset; const targetOffset = $(anchorId).offset().top - offset; window.scrollTo({ top: targetOffset, behavior: 'smooth' }); } }); $(this).after(jumpLink); });});
