A Microsoft spokesperson is typing something into a search engine and it doesn’t quite work. I’m watching this unfold at a Microsoft press event in Manhattan that’s meant to showcase the new features on Bing, the company’s Google rival. In this demonstration, a chatbot would answer a user question with an embedded video. Typing on a large computer monitor in full view of several reporters, the staff asks the program for instructions on how to tie a tie. But instead of a video, Bing generates an absurd pile of text, so many words about wrapping and knotting fabric on a sterile white speech bubble. It reminds me of a Times New Roman resource page you’d see on an old professor’s website.
Everyone in the crowd acknowledges that this outing is pointless. (Tie-a-tie.net, the oldest tie-related resource I can find on the web, also knew the score in 2003: Its how-to pages for styles like Windsor and Pratt had illustrations.) Another Microsoft rep makes a joke about how the glitch proves the point, it would be really helpful for the AI to show a video in this particular context and move on. They try something else and it works: The Bing bot gives a short answer to a question about skiing and then pops a YouTube video into the chat bubble.
It’s functional, if not inspiring, almost a mismatch for the luxe staging at the Microsoft Experience Center in Midtown, where reporters gathered yesterday amidst expensive-looking spring bouquets (the lilies smelled fantastic) and unlimited access to a free smoothie bar. The tech company, which has taken the lead in the race for generative AI, was excited to unveil what it calls the next wave of AI innovation. His vision is to transform the way people gather information and learn things from the internet. In immediate terms, that means opening up the Bings chatbot to anyone with a Microsoft account today; incorporate new types of media into search, such as video embeds and visual graphics; plug-ins that will allow a service like OpenTable to operate within the chat platform; and more. The juice flows.
Microsoft says this is the future of search. There’s been a lot of talk like this since last November, when OpenAI released ChatGPT and it seemed to turn the world upside down: a new generation of AI is suddenly more capable and, more importantly, more accessible than many would have thought possible. (Microsoft has invested billions in OpenAI and is using the company’s technology at Bing.) Each new day brings with it a different angle through which to view the prism: Maybe chatbots will help us at work, precipitate us more creative mixologists and/or redefine the nature of nuclear warfare. When it comes to research in particular, however, chatbots might just be unimaginative. My conclusion from seeing Bing in action was that AI-powered search was likely to expand the reach of human knowledge and take us to new frontiers online. Instead, Microsoft has squeezed AI into productivity software that makes the Internet seem smaller.
The problem, in a nutshell, is consolidation, a new twist on a problem that has plagued the internet for the last decade and a half or so, as social media giants, cloud providers and, well, Google have seized the benefits market and the absence of meaningful regulation to dominate our web experiences. Four years ago, journalist Kashmir Hill found it almost impossible to cut Microsoft, Amazon, Apple, Google and Facebook services out of her life. She thinks how much time spent on your phone is filtered through the same few services every day. You might still use a handful of different websites and apps, but less than the seemingly limitless scope of the Internet might suggest.
The Bing bot, along with ChatGPT, Google’s Bard and several competitors, portends a more drastic streamlining. Imagine that every crayon in the world has been melted in a dark globe and pinched through a funnel. Where you once went to a search engine to find another website to go to, you will now go to a search engine and stay on that search engine. For example, let’s say you go to a typical non-chatbot search engine like Ask.com and type in an everyday query like How do I clean mud off my leather shoes? I will receive a list of links, from various outlets and perspectives, and I will click on one of those links to hopefully find my answer. But now I can bring up the Bing chatbot and type the same thing; will present a six-step answer online, no external navigation required. Bing mentions links, but the entire product is designed to give you an answer within its chat interface. This is, clearly, the strong point.
I pitched the idea that the Bings chatbot could make the internet seem smaller during a quick interview with Yusuf Mehdi, corporate vice president and chief marketing officer at Microsoft. He called the product a co-pilot, something that could help people who cumulatively perform 10 billion Internet search queries a day. This articulation is significant: a co-pilot is essential. You wouldn’t want to take a flight without one. And on the Internet, the essentials take root. Once upon a time there was no Facebook, Instagram, Google or iCloud; now, for many, it’s hard to imagine life, let alone the Internet, without them. Digital technology is often positioned by companies in terms of expanding possibilities, but the net effect is to limit them. When asked if the new Bing was designed to keep you on Bing, rather than wandering off, Mehdi said he sees the issue as a potential risk, but fundamentally believes the Bing chatbot will be something of a liberating force, freeing people from the time consuming traditional search process as it is. What else can I learn about the world? What else can I go see? We’re just trying to take out a lot of the menial work of what people are doing and speed them up to get what they want, he said.
He also seemed to sense that there was an elephant in the room: namely, that I work as a journalist at a particularly brutal time for online media. Bing and other chatbot-based search engines could pose a threat to publications if search platforms discourage people from clicking on the original stories from which the information is taken. For us, it’s absolutely a goal to drive more traffic to content publishers, no questions asked, Mehdi told me. For example, it’s in our metrics internally. He explained the rationale: Publishers need clicks to sell ads, and the chatbot needs content from publishers to deliver anything to users. In some ways that’s true, in some ways it’s not: by its very nature, AI has been trained on so much existing material online that most searches outside of breaking news events are already well covered. I probably wouldn’t pour money into a new website about how to tie a tie.
I pushed again. Why is this goal important to you? He replied: It is important for us that the traffic is there, that it works, that publishers say: Yes, we like Bing; we like Bing chat. It’s getting us traffic; it is giving us volume. Then came another story. Ideally, publishers would be so impressed with the Bing chatbot that they want to integrate their services with it. As we talked about today, we want to have plug-ins, he said. We’d like people to build plugins on this.
And that made more sense. You may be waiting for traffic to come to your site. Or you could just build something that fits in the little white text box of the machine. After all, it’s the future.