[ad_1]
Everybody in the tech world, and even outside it, is talking about generative AI today. That’s also what we discussed with Brad Smith, Microsoft’s VC & president, who is in India to be part of the G20 dialogue forum with the global business community. Smith joined Microsoft 30 years ago and today, he’s representing it publicly on critical issues involving the intersection of technology and society. Microsoft is at the centre of the gen AI world with its massive investment in OpenAI — the non-profit that developed ChatGPT. Excerpts:
Is it gen AI that excites you the most right now?
It’s fair to say that this is the year of AI, and gen AI is obviously the focal point. It’s one of the years that I think will stand out historically as an inflection point for the history of technology and the tech sector. Gen AI will probably be more transformative for people’s lives than any other invention of our lifetime.
Give us a sense of this transformation you expect.
After ChatGPT first came out, we were talking about if it’s like the iPhone or the internet or electricity… we’ve had time to reflect on it and my view is that it is probably most similar to the printing press.
The printing press was perfected by Gutenberg in 1462 with movable type, and it made knowledge accessible to more people. Similarly, I think gen AI is a tool to help people learn and research. I experience that myself when I use something like Bing chat or what we’re now developing — the M365 Copilot. It’s a tool that one can use to create, to write more quickly, to turn a Word document into a PowerPoint deck and vice versa. It’s a tool that people who write code for a living are using to write more of their code. But we call it a co-pilot, not an autopilot… you need to continue to think. Just like the printing press, it can copy books but you are still needed to write and read them. And when one puts it in those terms, you start to see how it can make its way into every area of knowledge work that we have.
There are concerns that gen AI involves huge costs & energy consumption, that OpenAI is bleeding…
We’re still in relatively early days. I think we will see largescale models that, in effect, can be used to do many things. GPT-4 is a good example of that. I think you’ll see other narrower models, including open-source models, that probably won’t do as many things well, but they may do one or two things as well as a largescale model. Not everything is going to be larger and there’s a lot of innovation in our future to try to understand how these things come together.
There are concerns about the damage AI could do in the absence of regulations
We need companies to create own guardrails, and that’s what we’ve been doing for seven years now. We need to strengthen and add resources to the work we’re doing on something like, say, red teaming, so that when we’re working with OpenAI on the development of a new model, there is an independent red team at work to look at potentially sensitive uses, identify potential problems, and put in place measures to protect against them. Similarly, for each application — put a red team to work and identify the kinds of problems they can result in. Then you fundamentally create an AI safety architecture.
Is it gen AI that excites you the most right now?
It’s fair to say that this is the year of AI, and gen AI is obviously the focal point. It’s one of the years that I think will stand out historically as an inflection point for the history of technology and the tech sector. Gen AI will probably be more transformative for people’s lives than any other invention of our lifetime.
Give us a sense of this transformation you expect.
After ChatGPT first came out, we were talking about if it’s like the iPhone or the internet or electricity… we’ve had time to reflect on it and my view is that it is probably most similar to the printing press.
The printing press was perfected by Gutenberg in 1462 with movable type, and it made knowledge accessible to more people. Similarly, I think gen AI is a tool to help people learn and research. I experience that myself when I use something like Bing chat or what we’re now developing — the M365 Copilot. It’s a tool that one can use to create, to write more quickly, to turn a Word document into a PowerPoint deck and vice versa. It’s a tool that people who write code for a living are using to write more of their code. But we call it a co-pilot, not an autopilot… you need to continue to think. Just like the printing press, it can copy books but you are still needed to write and read them. And when one puts it in those terms, you start to see how it can make its way into every area of knowledge work that we have.
There are concerns that gen AI involves huge costs & energy consumption, that OpenAI is bleeding…
We’re still in relatively early days. I think we will see largescale models that, in effect, can be used to do many things. GPT-4 is a good example of that. I think you’ll see other narrower models, including open-source models, that probably won’t do as many things well, but they may do one or two things as well as a largescale model. Not everything is going to be larger and there’s a lot of innovation in our future to try to understand how these things come together.
There are concerns about the damage AI could do in the absence of regulations
We need companies to create own guardrails, and that’s what we’ve been doing for seven years now. We need to strengthen and add resources to the work we’re doing on something like, say, red teaming, so that when we’re working with OpenAI on the development of a new model, there is an independent red team at work to look at potentially sensitive uses, identify potential problems, and put in place measures to protect against them. Similarly, for each application — put a red team to work and identify the kinds of problems they can result in. Then you fundamentally create an AI safety architecture.
#Gen #AIs #impact #big #printing #press
