This analyst agrees with most points in the recent WSJ article (https://on.wsj.com/3X7dFts) with opposite conclusions. Most points made in the article have been repeated by us for months. Yes, GenAI in its current form is unsustainable, but the revolution is already jumping tracks to Robotics and a myriad of other fields updated our website (https://geninnov.ai/articles/).

1. “The rate of improvement is slowing”: As we have explained in articles describing the “Super Moore” pace, the model size could not keep doubling every few months without exhausting the Internet. However, the saturation is nigh in text-based data. The transformers have the ability to learn from vision and other fields with the same method. LLMs and their use cases may be saturated in text-based utilities like writing poems or sales pitches, but it is moving on to altering quantum computing, drug discovery, diagnostics, weather micro-forecasting, or material sciences.

2. “AI could become a commodity”: It has actually been that forever. Since the first articles a year ago, we have shown how building an LLM is one of the easiest things undertaken by hundreds globally with commoditization implications. But, this has only shifted business moats away from companies with only ideas to others we dubbed as with “the last mile.” There are multiple articles in our archives for those interested.

3. “Today’s AI remains ruinously expensive to run”: GenAI building is easy, but inference is expensive. As a result, we concluded that inference must move from cloud to consumer/corporate “pockets.” GenAI on the cloud with an XaaS model is also causing many unsavory business practices, which will likely cause many investigations when the downturn arrives. However, GenAI could spark a massive hardware replacement cycle; the green shoots of these are already visible.

4. “Narrow use cases, slow adoption”: The reality is different. LLMs and chatboxes are transient. They are already giving rise to Agents on one side and creating innovations across domains on the other. The Dollars, the human resources, and machines’ own exponentially rising abilities going into innovation are constantly creating new waves, ushering us into a global innovation race.

If GenAI is seen as nothing but a tool to search/query, code, or process text through computing on the cloud and revenues through subscriptions, there are massive disappointments around. There will also be disappointments for those who expect a moon through terms like AGI. It is a field where nothing is given, and the failure rates for individual attempts will be extraordinarily high – partly because of the results of innovation attempts but more because of hyper-competitiveness and continuous changes. But the GenAI world is already materially different from pre-GenAI, with all significant previous tech business models upstaged. Likely, we have barely begun.