AI Top-of-Mind for 6.12.24 — Software vendor switching
Top-of-mind is McKinsey’s latest research on the impact of Gen AI in software development and the acceleration of adoption. From the report:
· Gen AI will drive significant growth in the software space. By 2027, spending on the technology could reach between $175 billion and $250 billion, contributing an additional two to six percentage points of growth for the sector.
· However, despite that sizeable boost, our research suggests the most lasting and disruptive impact of gen AI will be a wide-scale acceleration of vendor switching, on the order of five to ten percentage points.
Opening the Stanford AI Index Report again, I’ll now turn to the chapter on policy. We all know of the focus on regulation both at the national and the state level. For the US, as an example, there is movement within the different agencies such as transportation, energy, and health. A good chart on the state of play in the US: Lack of movement in places like Pennsylvania and even Oregon is interesting.
Then at the national level, who has a strategy?
Sticking to policy, companies like Adobe, Disney, and Instagram have been under fire for their lack of transparency in the use of AI including use of user-generated content. This is becoming an issue as more firms license their content to LLM developers. ‘Creative Bloq’ discusses this issue and includes a clarification from Adobe that attempts to stem the concern:
Following the same thought, be prepared for content sharing by Meta. As ‘Tech Times,’ notes, they’ve announced that they will begin to share user content but not private messages to train their AI models. From the article:
· Beginning on June 26, Meta will begin using user data dating back to 2007 to train and enhance their AI technologies as part of the new privacy policy. Except for private messages, this covers user postings, images, captions, and messages to Meta’s AI chatbot.
· AI behemoths like OpenAI are taking a unique approach to model training in response to the data shortage issue. For example, OpenAI, the company that created ChatGPT, is reportedly considering training its GPT-5 model with transcriptions of publicly accessible YouTube videos. These strategies have been criticized, though, and video content creators may even file legal challenges in response.
Another look behind the curtain. I was forwarded a good posting on Meta’s AI data center architecture, with a diagram below, and if you want to dig deeper, there are also overviews for connectivity and networking.
