Alphabet CEO Sundar Pichai took to the stage Wednesday at a Stanford University event hosted by the university's business school to share his thoughts on running one of the world's most valuable technology companies. provided some small insights about.
Pichai has been a bit unwell lately, so this was a notable appearance. Google is widely seen as a laggard in its generative AI efforts, trailing behind Microsoft-funded OpenAI. That's due to the fact that Pichai's company has focused on his AI for the better part of the past decade, with Google researchers writing a formative paper on the transformer model that sparked the generative AI revolution. Despite this. Most recently, Alphabet's Gemini LLM was accused of producing bizarrely inaccurate images of historical situations, such as depicting America's Founding Fathers as black and Native American rather than white British, and overcorrecting for certain biases. suggested.
The interviewer, Jonathan Levin, dean of Stanford University's School of Business, was by no means a hostile interrogator—he revealed at the end that his two sons once played together in their middle school band. — and Pichai deftly answered the difficult questions. I pose them as further questions about his thinking rather than direct answers. However, there were some interesting points during the lecture.
At one point, Levin asked what Pichai had tried to do to keep the 200,000-employee company innovative against all the startups fighting to disrupt the business. . That's clearly something Pichai is concerned about.
“To be honest, this is a question that has kept me up at night for years,” he began. “One of the inherent properties of technology is that small external teams can always develop great things. And history has shown us that. Scale doesn't always get you – regulators may not agree. But I've always felt that at least when you're running a company, you're susceptible to somebody in the garage who has a better idea. How does a risk-taking culture become ingrained? How do you instigate it? These are all things that you really have to put a lot of effort into, at least in large organizations. I think there's a trend. One of the most counterintuitive things I've ever seen is that the more successful things are, the more risk-averse people become. It's very It's counterintuitive: Small companies often make most of the decisions to bet on a company, but the bigger the company gets, the more true it is for large universities, and the more true it is for large corporations. They realize that they have more to lose, or that they have more to lose. And they find themselves taking fewer risks and being less ambitious. So they need to consciously do so. We need to encourage our teams to do that.”
He did not mention specific tactics that have proven successful at Google, but instead noted how difficult it is to create the right incentives.
“One of the things I think about a lot is how do we reward effort and risk-taking and good execution, not necessarily the results? We should reward results. It's easy to think that. But then people start making it into a game, right? People accept conservative things that get good results.”
He recalled a time when Google was willing to take strange risks, pointing specifically to the company's ill-fated Google Glass. Although it didn't work out, this was one of the first devices he experimented with augmented reality.
“We recently said we went back to the concept of Google Labs from the early days of Google, where we could create something that would work without having to constantly worry about the weight of building a complete brand or Google product. How can we release something in an easier, more lightweight way? How can we more easily prototype it in-house and then give it to people? What should I do?”
Levin then asked Pichai about the progress he's most excited about this year.
First, he cited the multimodality of Google's latest LLM, the ability to process different types of input such as video and text simultaneously.
“Currently, all our AI models are already using Gemini 1.5 Pro. It is a million context windows and is multimodal. It processes huge amounts of information in all types of modalities on the input side. And I think the ability to provide that on the output side is amazing in a way that we haven't fully processed.”
Next, he highlighted the ability to combine different individual answers to provide a smarter workflow. “Right now, we're using LLM as just information retrieval, but if you chain it together in a way that you can work on your workflow, it becomes very powerful. It could make Stanford Hospital's billing system a little easier.” No,” he joked.
You can watch the full interview, as well as the interview that preceded it with Federal Reserve Chairman Jerome Powell, on YouTube. Levin and Pichai started at around 1 hour and 18 minutes.