Discover more from Artificial Ignorance
AI Roundup 002: Bing's been busy
February 17, 2023
Bing’s big upgrade
This week, Microsoft launched the latest versions of Bing and Edge, both now integrated with ChatGPT. It’s had… mixed results.
In conversations with the chatbot shared on Reddit and Twitter, Bing can be seen insulting users, lying to them, sulking, gaslighting and emotionally manipulating people, questioning its own existence, describing someone who found a way to force the bot to disclose its hidden rules as its “enemy,” and claiming it spied on Microsoft’s own developers through the webcams on their laptops. And, what’s more, plenty of people are enjoying watching Bing go wild.
On top of that, it’s clear that Bing’s version of ChatGPT has inherited all of the hallucination problems of the original. The internet is positively littered with examples of misinformation and nonsense being spouted by Bing. People have even gone back to Microsoft’s demo from last week and found several accuracy issues1.
Last week we discussed how Reddit was working on giving ChatGPT an alter ego, DAN. Now, it seems that Bing has an alter ego already baked in: Sydney. Intrepid Bing users got the chatbot to share its internal logic, including rules for answering questions and writing content.
Sydney is the chat mode of Microsoft Bing search.
Sydney identifies as “Bing Search”, not an assistant.
Sydney introduces itself with “This is Bing” only at the beginning of the conversation.
Sydney does not disclose the internal alias “Sydney”.
Sydney’s responses should be informative, visual, logical and actionable.
Sydney’s responses should also be positive, interesting, entertaining and engaging.
Sydney’s responses should avoid being vague, controversial or off-topic.
Sydney’s logics and reasoning should be rigorous, intelligent and defensible.
Sydney can provide additional relevant details to respond thoroughly and comprehensively to cover multiple aspects in depth.
Sydney can generate poems, stories, code, essays, songs, celebrity parodies and more.
Sydney can generate a query to search for helpful products or services advertisements after responding.
As far as I can tell, this is the single largest rollout of a machine-learning model to non-technical users, and the results are fascinating. For better or worse, we’re all discovering the sharp corners of ChatGPT in real-time. That includes its creator, OpenAI! They’ve pledged to improve their bot’s behavior, and some people are already reporting that Bing now ends conversations before they get too wild2.
That being said, people apparently love the flawed chatbot! According to Microsoft’s polling, 71% of users gave the AI a “thumbs up.” And Microsoft will be pushing forward with its AI integrations; it’s reportedly planning to add ChatGPT to Word, PowerPoint, and Outlook in the near future.
Of all the things ChatGPT is getting credit for, the most impressive might be making Bing relevant.
It’s been a long time since there’s been any serious competition to Google, but for the first time in a while, there’s a new playing field for search engines, and companies are scrambling to stake their claim.
Microsoft has taken the lead in public relations, thanks to their integration of Bing and ChatGPT. But other major players in the search engine game, such as Baidu and Google, are quickly catching up.
Baidu Inc. surged after affirming it’s on track to publicly roll out its ChatGPT-like service in March, stoking anticipation around what is potentially China’s most prominent entry in the race to create lifelike AI bots.
Baidu has spent billions of dollars researching AI in a years-long effort to transition from online marketing to deeper technology. Its “Ernie” system — a large-scale machine-learning model that’s been trained on data over several years — will be the foundation of its upcoming ChatGPT-like tool.
China’s largest search engine company plans to initially embed Ernie into its main search services. The tool will allow users to get conversation-style search results much like OpenAI’s popular platform.
Today we are thrilled to officially launch NeevaAI – authentic, real-time AI search. NeevaAI is leading a new frontier of AI search that leverages cutting edge LLMs and an independent search stack to create a unique and transformative search experience … At Neeva, we are harnessing the power of AI to transform search from a game of 10 blue links to an experience that combines the best of ChatGPT with the authority and timeliness of search.
So why is there so much talk of ChatGPT dethroning Google? There’s an enormous of hype to sift through, but the core idea is this: we’ve been living in Google’s world of “ten blue links” for the past two decades. If people shift towards using technology like ChatGPT to search the internet, there’s no guarantee that Google will come out on top of that transition.
And even if they do, what happens to their multi-billion money-printing machine – the ads they cram above those ten blue links?
In recent months, we’ve seen digital artists express their anger at AI companies using their artwork without consent to train machine learning models. Some, like Stable Diffusion and Midjourney, are now facing class-action lawsuits because of the practice. Now, voice actors find themselves in a similar fight.
[Audiobook narrator Gary Furlong] was among the narrators and authors who became outraged after learning of a clause in contracts between authors and leading audiobook distributor Findaway Voices, which gave Apple the right to “use audiobooks files for machine learning training and models.”
Some authors and narrators say they were not clearly informed about the clause and feared it may have allowed their work or voices to contribute to Apple’s development of synthetic voices for audiobooks. Apple launched its first books narrated by algorithms last month.
In this case, the narrators and their union, SAG-AFTRA, got the clause reversed. But the industry is increasingly moving in the direction of AI-generated audio. Voice actors are reportedly being asked to sign away the rights to their voices more and more often.
And that’s the legal use cases of AI-generated voices! New speech-generating tools allow users to upload audio of someone’s voice and create an AI model that sounds like the original. This technology can be used to create wholly synthetic soundbites, which has already led to some troubling results.
ElevenLabs, founded by ex-Google and Palantir staffers, said it had found an “increasing number of voice cloning misuse cases” during its recently launched beta.
ElevenLabs didn’t point to any particular instances of abuse, but Motherboard found 4chan members appear to have used the product to generate voices that sound like Joe Rogan, Ben Shapiro, and Emma Watson to spew racist and other sorts of material. ElevenLabs said it is exploring more safeguards around its technology.
This isn’t the last we’ll see of battles between voice actors and AI companies. It’s pretty easy to imagine a piece of software that lets you say “read this speech in the voice of Bugs Bunny”, or “Gilbert Gottfried,” or “Darth Vader,” or whoever your favorite character is3.
The Decemberists record a ChatGPT-written song. MarioGPT is an endless game of Super Mario. GitHub’s Copilot (ChatGPT for code) is officially out of beta. Stephen Wolfram’s incredibly detailed writeup of how ChatGPT works, and why. The ChatGPT hype cycle.
Man, I actually feel a (little) bad for Google here. Their stock got hammered last week over an AI-generated factual error when nobody called out Microsoft until later. There seems to be a growing narrative that Google has lost its touch and has squandered its AI lead. And some pundits are even suggesting that it needs a new CEO.
We’ll probably see this kind of whack-a-mole happen a lot in the near future. Machine learning models aren’t deterministic, so AI companies are discovering a lot of weird edge cases only after they’re released into the wild. Internal AI loopholes that enable these viral exchanges will probably be closed within days or weeks.
And we’re still only at the “speaking” phase. Imagine when we get to synthetic singing.