Threads, ChatGPT usage drops, and AI demos with Sunny Madra | E1774

AI Usage

Jason Calacanis and Sunny Madra explore the recent decline in ChatGPT's growth, emphasizing that this trend doesn't necessarily indicate waning interest in AI. Sunny argues that many AI use cases have been integrated into other products, making direct usage metrics less reflective of true engagement. He compares this to AWS, where initial usage was primarily internal before expanding to external clients.

People aren't less interested in it, which is such a stupid take because obviously when a new technology comes out, everybody tries it because there's no cost to trying it.

--- Jason Calacanis

Jason adds that the real measure of AI's impact is in developer engagement and the subsequent consumer applications that emerge from it 1 2.

   

Model Training

The conversation shifts to the challenges of training new AI models, particularly in acquiring unrestricted data. Sunny highlights the complexities of data storage and the inability to reverse-engineer language models to extract original data. He explains that data is transformed into embeddings, making it difficult to untangle specific information.

If you've ever trained a model and you have some type of restricted data in there, it is in the model forever until you retrain.

--- Sunny Madra

Jason questions whether future models will be weaker due to these restrictions, and Sunny suggests that starting from scratch could indeed pose challenges 3 4.

   

AI Investments

Sunny and Jason discuss the massive investments in AI infrastructure, with a focus on Inflection AI's significant hardware spending. Inflection AI, backed by notable investors like Microsoft and Nvidia, plans to build a large cluster of AI systems, highlighting the strategic importance of hardware in AI development.

Inflection AI, it started by co-founder of DeepMind Mustafa and Reed Hoffman, and they raised a massive amount of money, I think $1.3 billion on a significant valuation.

--- Sunny Madra

Jason speculates on the potential stock market implications of these investments, suggesting that such deals could inadvertently boost the stock values of involved companies 5 6.