If you are using Big AI Chat you may want to download Microsoft Dev - that is their latest beta release - you can have both Edge and Edge Dev on your computer at the same time
The current Dev version gives you a feature shown in the Microsoft live demo on Youtube but puzzlingly not available in the standard Edge release. It summarizes the current page and makes it very easy to ask Chat questions or compose content related to the website. This is a killer feature for me when doing research on the web and may be the final straw that makes me give up Chrome entirely.
Features in Edge Dev take about a month to propagate to the Beta channel, and then some more time to be included in Release, i.e. Edge Dev has features that the standard release will have included in a couple of months (Canary → Dev → Beta → Release).
Do you mean this “Discovery”-Window?
Why do you think this is a killer feature (other than it might kill the application at all )?
It has three points, with an almost identical content, while the other important information from the article, as far as it is visible on the picture, are not even mentioned.
How could anybody do a thorough research with a (dis)function like that?
Because you can be infinitely specific and creative in the prompts you use - such as:
Detailed Summary of main web page
4 bullet point summary
What are links to 5 similar articles?
What are links to 5 articles with opposing views?
Show me competitors to this product
What televisions are similar to this but bigger
Show me yelp reviews for this place
Also you can switch to a different web page but keep the prompt window intact and then ask the same questions about the new web page - or ask for a comparison between the pages. Or a chart comparing the pages, with whatever columns you request.
No doubt more creative/useful use cases exist for both personal and professional situations
Yes this is a killer feature
Another helpful way to use it - highlight any text on a web page and copy it to the clipboard. That text automatically shows up in the Discovery window and you can then ask a question about it.
But what is your advantage, if all of your bullet Points contain the same content, but did not cover the whole article, as demonstrated in your screenshot?
And what could be the advantage, if it is proven, that the system is producing false results, and is simply lying to you? How could that be a help of any kind!
It is the other way around, as you not only not get useful/thrustful results, but you have to spent a lot of extra time, to check whether a result might be correct, or not. This is, in my opinion, a huge waste of time!
For what it’s worth - hallucination is indeed a major risk of GPT. However that almost always happens when you do a searach and it cannot find the answer.
On the other hand, one of GPT’s biggest strengths is the ability to summarize information. If you ask it a question about what is on a given page, the answer may or may not be accurate depending on the situation. But if you simply ask to to “summarize main page” it is extremely capable in that regard. Similarly it is very good at finding additional sources which support or rebut a given article.
For those purposes -which reflect a big part of my daily activity - Bing AI is extremely effective. Your use case may well be different.
As I read more about the statistical way words are chosen by an AI chat bot for its replies (with some random variation thrown in to make things sound less flat), I can see how a text might seem as if it had been written by a person and might even seem useful.
But for a text to actually be correct and useful apparently involves a lot of luck, and then a lot of verification by the person conversing with the chat bot.
However if you present it with a long document and ask it to summarize, find other supporting sources, or find other sources with alternative views on the topic, then its performance is consistently A+.
Isn’t this (asking a random question) what most people are doing when they converse with something billed as an Artificial Intelligence-Powered Chat Bot?
Your Screenshot above, proofs you wrong.
ChatGPT is simply not doing, what you think it is. It seems to do the thinks you want it to do, but in truth, it simply writes one word after the other, without any intend of a useful content.
The system is not knowing, that you want it to do the specific tasks you are expecting from it, and while it do not know it, it couldn’t do it.
Yes, of course you could answer questions, and you will get answers and results, but they do not have the meaning, you want them to have.
If you do that for fun, go for it, but if you are doing it for some Business stuff, I would be very careful…
Yes - many are doing that. That’s not my use case. I am a bit skeptical whether that use case will last long-term. If yes, good for Microsoft. Not my cup of tea for sure.
I’m glad you have found a use for the current crop of chat bots. You are a knowledgeable user who is aware of the pitfalls to look out for. I’m concerned for the “normies” who won’t know any better: chat bots landed on the the front page of my local newspaper yesterday (via syndication from the New York Times) in an article entitled “An unsettling conversation with Bing.”
)?