As I assemble today’s show, Apple is launching stuff at their WWDC show. In preparation, I’m clearing the decks of my AI backlog, so expect Apple news in a subsequent episode.
According to new Pew Research Data, about six-in-ten U.S. adults (58%) are familiar with ChatGPT, and 14% of US adults have tried ChatGPT. And in a new Morning Consult study done in April, 38% of respondents indicated they “trust artificial AI companies to do the right thing,” while 26% said they didn’t know and 36% indicated they “don’t trust” AI companies. There are generational and political divides in the data too. Younger and big-D Democratic are more likely to trust the tech.
Last week, OpenAI announced a new method for training models designed to help reduce AI hallucinations. The new strategy, quoting CNBC: Train AI models to reward themselves for each individual correct step of reasoning when arriving at an answer, instead of just rewarding a correct conclusion. The approach is called “process supervision,” as opposed to “outcome supervision,” and could lead to better explainable AI, according to the researchers, since the strategy encourages models to follow more of a human-like chain of “thought” approach.
A notable use – Artifact’s new application can now re-write click-bait style headlines in real-time, using OpenAI. If enough users mark a story as clickbaity, the application might start using an AI-re-written headline for all users.
The Information has insights into the fees some of those Office 365 customers are paying to test the new AI features. From the piece. At least 100 customers are already paying a flat fee of $100,000 for up to 1,000 users for one year, according to a person with direct knowledge of the pilot program. That means the customers with 1,000-seat subscriptions are paying at least 40% more versus the classic version for productivity software that automates tasks such as writing text in Word documents and creating PowerPoint slides.
A local but representative story – Arlington County, Virginia’s non-emergency 911 line will be answered by an Amazon-powered artificial intelligence (AI) device. This started June 1, and the intention is to reduce the load to critical 911 calls.
Despite all this, add another company to the list restricting employee use of the tech. iHeartMedia joins the list. Then again, Next Web looks at Unilever’s use of AI technology. Accelerating lab work, product design, and supply chain lead the list of the case study I’ve linked to.
A big thought here, too – Axios is covering the discussion of a change in computer interface design due to AI. Is AI the next giant leap in interface design?
Much of the headlines around the 14% of adults implied this was low. I disagree – 14% of the country has tried this technology, and it’s only been out for six months. That’s remarkable. Technological shifts are far slower than our perception thinks they should be.
Another thought here – there are a lot of older men running IT services companies. If you’re in that category, be self-aware enough to know that your customer and user base may not look like you.
I included several stories about use cases to open the creative mind up for applications for customers. Think about an application now re-writing data before being displayed to users. There’s something both very cool and hazardous to that. Synthesizing information for me is powerful; obfuscating it from view is dangerous.