OpenAI's GPT-4.5 Could Be a Canary in the Coal Mine for the AI Bubble

The signs that large-language models are starting to hit a ceiling in capabilities regardless of how much money and computing resources are thrown at them are piling up. Late last year, the founders of venture capitalist firm Andreessen Horowitz noted in an interview that the gains delivered by each successive generation of AI models were getting smaller. "They're sort of hitting the same ceiling on capabilities," Andreessen said about the various companies working on advanced AI models.Part of the problem is data. At this point, the best AI models are already trained on essentially all the digital data available. Without more data, improvements in capabilities will have to be driven by new training methods or other innovations.OpenAI kicked off the AI boom in late 2022 with ChatGPT, which was powered by the company's GPT-3.5 model. GPT-4 soon followed, which marked a major improvement in capabilities, and the company later launched a series of other models as part of the GPT-4 family.Continue readingWeiter zum vollständigen Artikel bei MotleyFool Weiter zum vollständigen Artikel bei MotleyFool