Everybody though that Google was behind and failing when their problem was to understand how far of being a production product this kind of chats are.
Everybody though that Google was behind and failing when their problem was to understand how far of being a production product this kind of chats are.
I had a 7 response back and forth with it today trying to establish the American Liberal English equivalent of “I am a Patriot” from American Conservative English. Each of my responses was why I disagreed with its results and what it could do better.
Being able to type five sentences in a row, before submitting anything, and get a coherent cited response is utterly unlike googling some words and opening the top 10 results hunting for something.
That is why Bing is the wrong product to integrate it with. Most Google users will not look after the first 10 results because reading so much gets tiresome. They are never going to write 2000 words to try to get the information that they want when they are not willing to read 100 words.
> I disagreed with its results and what it could do better.
That is the problem. You know the answer and are looking for a question to get it. That is the opposite of what a search engine is. It can be a great tech for makes no sense as a search engine.
We sort of settled upon there not being a great counterpart.
I ended up with a 500 word query because of the question limit. I initially asked it a question with 4 follow ups. After pressing new topic, I rephrased all my previous query as one request, and it handled my previous questions in one response.
Simulating reasoning is only an impressive trick if you are accurate.
ELIZA "simulated reasoning" - but it's just a pattern matcher. Now this bot completes sentences based on sentences that other people have emitted in the past.
> get a coherent cited response
It would much more interesting if it were an accurate response.
Who knows how long this stays in a prerelease state, and what other sanity checks they add to it.
The primary purpose seems to be to overstate the capabilities of these relatively incapable language models in order to manipulate investors and in particular to interfere with specific segments of the labor market in order to suppress salaries.
Otherwise, I've been seriously struggling to understand what value these easily broken and lying _chatbots_ would have.
It is not a hidden fact that CEOs wants to hype companies to increase share value. That is the basics of 2023 economy, hype over results.