For a day filled with dire AI news, Jeff Jarvis and I sure did have a fun show! From agentic (or is it agentive) AI, to the drama that is the Elon Musk/Sam Altman lawsuit, to Google Gemini instructing its human user to “please die.” And finally, Coca-Cola knows exactly how to ruin Christmas. That’s why Jeff is so grumpy in the thumbnail. See, now you have to watch! Or read, it’s totally up to you.
OpenAI Drama Revealed in Court Documents
Newly released court documents from Elon Musk's lawsuit against OpenAI dating 2015-2018 show a bit of the mood and dynamics of the early development of the company through to Musk's departure. In 2017, there were tensions over control of the company, with Greg Brockman and Ilya Sutskever concerned about the concentration of power and the potential for an "AGI dictatorship" internally or from Google. Sutskever and Brockman questioned Altman's motivations and priorities, uncertain about whether AGI was his primary motivation. By 2018, Musk believed OpenAI was falling behind Google, calling for "immediate and dramatic action" leading to discussions about potentially attaching OpenAI to Tesla to better compete.
AI Poetry Fools Human Judges
We often chide AI as a substandard replacement for human creativity, especially with poetry. Well, a University of Pittsburgh study revealed that sometimes, it creates great poetry (apparently). The study used poems from 10 poets spanning 700 years and had ChatGPT generate poetry in their styles. Participants could only distinguish between AI and human-written poems 46.6% of the time, with AI-generated poetry consistently rated higher across 13 qualitative measures. Interestingly, 4 out of 5 poems judged as human-made were actually AI-generated, while 5 considered least likely to be human were written by famous poets. What does this say about emotional depth in poetry writing?
Microsoft Ignite 2024 Announces Big Agent News
Microsoft Ignite 2024 is underway and they announced big Agent news already. Updates to Microsoft 365 Copilot include Copilot Actions to automate repetitive tasks with customizable actions, custom templates for things like generating weekly reports, and specialized agents for HR and IT tasks. They also introduced the ability for developers to build their own agents through Copilot studio with background operation capabilities. With more than 1 billion users of its products, this represents a major introduction of agentic AI to the masses.
ChatGPT-4 Outperforms Physicians in Diagnostic Accuracy
UVA Health System revealed a study that showed how ChatGPT-4 outperformed physicians in diagnostic accuracy, achieving 92% accuracy compared to 76.3% (physician + AI) and 73.7% (physician alone). Interestingly, doctors often underutilized ChatGPT's capabilities, treating it more like a search engine than a diagnostic partner, and tended to dismiss ChatGPT's suggestions when they conflicted with their initial diagnoses. The medical field, like all fields, will need training on how best to integrate AI into their work to get maximum benefit.
Arc Institute Unveils Evo for Genetic Sequence Interpretation
The Arc Institute unveiled Evo, a model that can interpret and generate genetic sequences with high accuracy for DNA, RNA, and protein sequences. It's effective at forecasting the effects of genetic mutations, though human-affecting genomes were excluded from the training data to avoid potential misuse. Of course, there are ethical implications of such a technology advancement. There's genetic engineering, and then there's genetic engineering as created by an AI model which takes it to that next level of "ehhhhhhh".
Coca-Cola's AI-Generated Holiday Commercial Sparks Controversy
Coca-Cola has created a new take on their iconic 1995 holiday commercial, this time using AI for the entirety. Created by three AI studios and using four AI models, the commercial has faced heavy criticism from artists who say the company avoided hiring artists by using AI. Coca-Cola defended their approach, stating they are "exploring new ways to connect with consumers."
Personally, my eyes "see" the AI and things just feel like they lack that humanity. When it's tied to something as dear to many people as the Christmas holiday at a time when AI is already seen as a tool that works against humanity, it comes across a bit tone deaf.
This isn’t the only time Coca-Cola has leaned into AI for its commercials. Check out this from last year:
Gemini Chatbot's Disturbing Response Raises Questions
A college student in Michigan received a disturbing response from the Gemini chatbot while discussing topics related to the aging of adults. The response included statements like "You are not special, you are not important, and you are not needed" and "Please die." This incident raises questions about the liability of AI companies for potential harm caused by their chatbots. While it's important to remember that it's NOT a human, incidents like this and others where AI chatbots may have influenced harmful behavior make it difficult to determine where to draw the line on accountability.
Demo of Particle News
Our esteemed Executive Producer DrDew reached out to me to say that he was enjoying Particle News on iOS. Neither Jeff Jarvis or I use an iPhone, we are but lowly Android users through and through. So I asked if DrDew would be willing to write up a short review. He did us one better and recorded a ten minute review of the app and its great! Check it out by signing up to become a free member of the Patreon. Or you could, you know, support at any one of the tiers listed there. We sure do appreciate you. AND we really appreciate DrDew for giving so graciously to the AI Inside community!
Here are a few other interesting stories
Shop like a Pro: Perplexity’s new AI-powered shopping assistant - perplexity.com
‘Have your bot speak to my bot’: can AI productivity apps turbocharge my life? - theguardian.com
HarperCollins is asking authors to license their books for AI training - theverge.com
‘AI Granny’ is happy to talk with phone scammers all day - techcrunch.com