The first section of yesterday’s LTT Wan Show has a section where they try out new bing. It is pretty impressive. For example, they ask how many LTT store backpacks fit in a trunk of a Tesla. Bing...
The first section of yesterday’s LTT Wan Show has a section where they try out new bing. It is pretty impressive.
For example, they ask how many LTT store backpacks fit in a trunk of a Tesla. Bing first searches the web for that directly, but can’t find it. So it searches for the dimensions of each and comes up with an estimate based on that. It hedges its answer by saying that both the trunk and backpack are weird shapes, and it will depend on the packing efficiency. Then , to find a possibly better answer, it comes to the conclusion that the backpack is a similar size to a suitcase, and finds and sources a video about how many suitcases fit in a Tesla trunk.
Later on, they ask “How do I defeat Biomass”, with no context. They wanted to see if it would realize that biomass is a boss from Valheim. But Luke forgot the name of the boss, which was actually bonemass (or something like that). Bing was able to make the connection anyway (along with a few other worrying answers) and asked for more clarification. Seriously impressive.
That gave me much more confidence in how useful it'll be in addition to demo's like on LTT. It seems like it does bing searches (maybe guided by the model - that "backpack in tesla" question was very involved), then feeds the information back into the LLM to summarize and answer.
That's also pretty damn sick. It's basically automating the search process; first you make some broad queries, read the pages that come up, start collecting that information in your head, do more searches, and finally synthesize the information into a conclusion in your head. It looks like it just does that for you. Which is actually very useful.
Announcement on the Microsoft blog: https://blogs.microsoft.com/blog/2023/02/07/reinventing-search-with-a-new-ai-powered-microsoft-bing-and-edge-your-copilot-for-the-web/
The Microsoft Edge dev and canary builds (not beta) have the edge features kinda enabled. They are there, but you still need to get through the waitlist to do anything with gpt. It does give some...
The Microsoft Edge dev and canary builds (not beta) have the edge features kinda enabled. They are there, but you still need to get through the waitlist to do anything with gpt. It does give some basic traffic metrics sourced from bing, which is cool.
Many searches people do are essentially questions (how to do ..., what is ..., etc.), having ChatGPT answer these may or may not be better than whatever results come up from other websites. I have...
Many searches people do are essentially questions (how to do ..., what is ..., etc.), having ChatGPT answer these may or may not be better than whatever results come up from other websites.
I have no interest in any of this so I don't know if Bing does it but being able to essentially have a conversation to get your answer, rather than having to refine isolated searches, I can imagine would be a game changer.
There's an art to using a search engine well, that involves a lot of prediction about what words/phrases you expect to find in your result content. For some searches this is straightforward, but...
There's an art to using a search engine well, that involves a lot of prediction about what words/phrases you expect to find in your result content. For some searches this is straightforward, but in many domains you can't get a lot of precision and increasingly the results are full of worthless blogspam. A savvy user will work around this; probably most users will be lost.
ChatGPT is shockingly good at sussing out your intent, even if you don't know the name of something or terminology related to it. I haven't used the new Bing integration but if it's done well it should be a huge leap forward in getting relevant answers to broad queries.
It's my understanding that it can also summarize result content in aggregate. This means it can remove the need to click and read through multiple sites looking for a needle in a haystack. This could be a huge time-saver. It may also result in sites removing all the nonsense filler "content" they added for SEO, since it's no longer needed. I'm not sure if that particular point is true but I can dream. I'd like to think the blogspam problem will go away since the AI can wade through the noise and find the signal faster than humans can... but I'm sure there will be a huge push to find new ways to game the system.
To add on to this: there seems to be a generational gap with how search engines are used. When I watch my students search things online, they almost always use full, natural language questions...
To add on to this: there seems to be a generational gap with how search engines are used. When I watch my students search things online, they almost always use full, natural language questions instead of keywords.
I would search something like “volume formula sphere”, but they would search something like “what is the formula for the volume of a sphere?”. Sometimes this is comical, as we’ll often see searches like “what is a good original sentence that shows an understanding of the meaning of ‘commensalism’?” (which is part of the reason why so many teachers are worried about tools like ChatGPT).
Also, they are VERY dependent on the snippet at the top of the search results that Google pulls. I have to actively train them to actually scroll down and click on search results, which they hate doing and consider a massive pain. Having something like ChatGPT that gives instant, accurate answers first thing would be huge for the way my students use search.
“New Bing” will, but it’s a private beta at the moment. I think it’s unwise to roll it out on a waitlist, when Google is racing to introduce their own competitor. This was newsworthy enough for...
I have no interest in any of this so I don't know if Bing does it
“New Bing” will, but it’s a private beta at the moment. I think it’s unwise to roll it out on a waitlist, when Google is racing to introduce their own competitor.
This was newsworthy enough for the New York Times weekend quiz to ask if people knew that Bing still existed.
The first section of yesterday’s LTT Wan Show has a section where they try out new bing. It is pretty impressive.
For example, they ask how many LTT store backpacks fit in a trunk of a Tesla. Bing first searches the web for that directly, but can’t find it. So it searches for the dimensions of each and comes up with an estimate based on that. It hedges its answer by saying that both the trunk and backpack are weird shapes, and it will depend on the packing efficiency. Then , to find a possibly better answer, it comes to the conclusion that the backpack is a similar size to a suitcase, and finds and sources a video about how many suitcases fit in a Tesla trunk.
Later on, they ask “How do I defeat Biomass”, with no context. They wanted to see if it would realize that biomass is a boss from Valheim. But Luke forgot the name of the boss, which was actually bonemass (or something like that). Bing was able to make the connection anyway (along with a few other worrying answers) and asked for more clarification. Seriously impressive.
There was also an interview on the NYT's Hard Forked with OpenAI executives on the bing integration (among other things): https://www.nytimes.com/2023/02/10/podcasts/bings-revenge-and-googles-ai-face-plant.html
That gave me much more confidence in how useful it'll be in addition to demo's like on LTT. It seems like it does bing searches (maybe guided by the model - that "backpack in tesla" question was very involved), then feeds the information back into the LLM to summarize and answer.
That's also pretty damn sick. It's basically automating the search process; first you make some broad queries, read the pages that come up, start collecting that information in your head, do more searches, and finally synthesize the information into a conclusion in your head. It looks like it just does that for you. Which is actually very useful.
Announcement on the Microsoft blog:
https://blogs.microsoft.com/blog/2023/02/07/reinventing-search-with-a-new-ai-powered-microsoft-bing-and-edge-your-copilot-for-the-web/
From the article:
I didn't see any sign of it on Bing other than a "Chat" tab that prompts you to join the waitlist. Which I did. We'll see.
The Microsoft Edge dev and canary builds (not beta) have the edge features kinda enabled. They are there, but you still need to get through the waitlist to do anything with gpt. It does give some basic traffic metrics sourced from bing, which is cool.
Well, it's a product of Microsoft. It's not supposed to work.
Can someone ELI5 how can a search engine benefit from something like ChatGPT?
Many searches people do are essentially questions (how to do ..., what is ..., etc.), having ChatGPT answer these may or may not be better than whatever results come up from other websites.
I have no interest in any of this so I don't know if Bing does it but being able to essentially have a conversation to get your answer, rather than having to refine isolated searches, I can imagine would be a game changer.
There's an art to using a search engine well, that involves a lot of prediction about what words/phrases you expect to find in your result content. For some searches this is straightforward, but in many domains you can't get a lot of precision and increasingly the results are full of worthless blogspam. A savvy user will work around this; probably most users will be lost.
ChatGPT is shockingly good at sussing out your intent, even if you don't know the name of something or terminology related to it. I haven't used the new Bing integration but if it's done well it should be a huge leap forward in getting relevant answers to broad queries.
It's my understanding that it can also summarize result content in aggregate. This means it can remove the need to click and read through multiple sites looking for a needle in a haystack. This could be a huge time-saver. It may also result in sites removing all the nonsense filler "content" they added for SEO, since it's no longer needed. I'm not sure if that particular point is true but I can dream. I'd like to think the blogspam problem will go away since the AI can wade through the noise and find the signal faster than humans can... but I'm sure there will be a huge push to find new ways to game the system.
To add on to this: there seems to be a generational gap with how search engines are used. When I watch my students search things online, they almost always use full, natural language questions instead of keywords.
I would search something like “volume formula sphere”, but they would search something like “what is the formula for the volume of a sphere?”. Sometimes this is comical, as we’ll often see searches like “what is a good original sentence that shows an understanding of the meaning of ‘commensalism’?” (which is part of the reason why so many teachers are worried about tools like ChatGPT).
Also, they are VERY dependent on the snippet at the top of the search results that Google pulls. I have to actively train them to actually scroll down and click on search results, which they hate doing and consider a massive pain. Having something like ChatGPT that gives instant, accurate answers first thing would be huge for the way my students use search.
It isn't need, but greed. Sites will eventually find ways to trick ChatGPT.
“New Bing” will, but it’s a private beta at the moment. I think it’s unwise to roll it out on a waitlist, when Google is racing to introduce their own competitor.
This was newsworthy enough for the New York Times weekend quiz to ask if people knew that Bing still existed.