Wired: When Openai launched Chatgpt, Google opted for a more cautious approach. Was it a deliberate strategy to prioritize security or technical differentiation? How do competition and ethics balance in AI?
Jeff Dean: Yes, it was definitely a deliberate strategy. In fact, we had internal chat systems similar to what Openai launched, but much earlier, maybe a year and a half. During the pandemic, many of our employees used them from home. At lunch, for example, you could see how the use of the system increased because people, boring at home, looked for someone to talk and talked with the system. However, we noticed that, although he had impressive abilities, he also presented problems: sometimes he said incorrect things with great confidence. Since then, we have advanced a lot. Part of our approach has been responsibly improve the capacities of these models and their behavior, aligning them with what we want to offer. It is not always perfect, but we prefer to be cautious: it is better to give users something safe and useful, even if that means that the model sometimes responds ‘I do not know’, instead of risking us to say something toxic or incorrect. For us, it is crucial to demonstrate that level of care and consideration towards those who use our products.
Wired: What is your vision of the impact of AI on the future?
Jeff Dean: It is an exciting area. Recently, together with eight co -authors, I published an article called Shaping AI where we explore how AI could affect sectors such as health, education, national security and employment. We believe that, if society manages this technology well, the benefits will be huge. It is not about seeing it as something inherently good or bad, but to mold it to maximize its positive aspects and mitigate the risks. For example, in health or education, AI could be transformative.
WIRED: What do you think of competition with companies such as OpenAi or China’s role in AI?
Jeff Dean: It is natural that multiple actors explore the AI. China has talented engineers and researchers, and it is logical that they also advance in this field. We focus on making our systems as useful as possible, without getting too distracted with what others do. Competition is healthy, but our priority is quality and ethics.
Wired: In a context of climatic crisis, how does the AI boom affect?
Jeff Dean: It is a critical issue. The increase in data centers and computational capacity (for AI and other uses) requires energy. Therefore, we prioritize efficiency: our TPUS consume less energy, and many centers, such as Chile, work almost 100% with renewable energy. In addition, some studies on the environmental impact of AI have serious errors (some exaggerate for a 100,000 factor!). We must base ourselves on real data and continue improving efficiency.
Wired: Where do you see a greater impact of AI in Latin America? Are there local projects that excite you?
Jeff Dean: In health and education. In regions with few doctors per inhabitant, AI can expand access to medical expertise. In education, our Notebook LM product allows you to convert texts into conversations or podcasts generated by AI, adapting to how each student learns. We also work in translation to languages such as Quechua, which democratizes access to information for historically marginalized communities.
#Jeff #Dean #architect #molds #future #Google #world
