“AI” is already handling the search for you. The big search engines are probably the first mass scale adopters of machine learning.
And they have lost the war with SEO spam to a hilarious extent. What makes you think the same won’t happen with chat bot AIs? Bad actors (including PR agencies) will inevitably figure out where and how to spam comments in order to bias the AI models in favor of their agendas or products.
If the data they consume is filled with something like “fossil fuels don’t cause global warming because XYZ”, the chat bots will repeat it. They don’t have the capacity to reason.
There hasn’t been a reason to flood the internet with low effort spam because it’s easily detected by humans who read it. But the ML algorithms will be a lot easier to trick.
Apologies, I used the overly vague term “AI”. Any company creating an LLM that has web search + scraping capabilities will be at the mercy of the search results.
That said, LLMs are actually quite skilled at ignoring noise (repetitive data), so gaming SEO may lose popularity. Hell, the practice will DEFINITELY lose appeal once LLMs are just browsing for relevant content and summarizing without any citations (links to the sites). And even of they do cite, no one will click them.
Convenience > Fact
tldr; This additional layer of obfuscation between search and result will reshape the fabric of the internet with time
You can already outsource a lot of this to Bing. If you need to know the right temperature for making french fries, you can google a bunch of “recipes” (AKA life story of the author + history + vacation photos + cooking instructions) read them through and… actually better make some coffee while you’re at it because this is going to take a while. Anyway, the other option is to ask: “Hey Bing, I’m making french fries, but I don’t know how hot the oven should be.”
Spoiler: 220 °C
The scary thing is, what happens when people start doing this for more important things, such as what to do if your child has swallowed something or how to parallel park your car.
Correct. However, if you buy frozen ones, you do need to heat them up some way. I ran out of nuclear weapons again, my flamer was out of gasoline, so using the oven was my best option.
It already is. If you want to play a game of D&D with chatGPT, there’s a very specific prompt carefully crafted for that. If you want to chat with a with a total psycho, there’s a prompt for that. If you want your AI to do something it was specifically forbidden from doing, just craft a very specific prompt for that, and you’re good to go. You can even find sites that collect various prompts for just about any purpose you can imagine.
Once AI is handling search for us, many may never learn the concept of “search term”
“AI” is already handling the search for you. The big search engines are probably the first mass scale adopters of machine learning.
And they have lost the war with SEO spam to a hilarious extent. What makes you think the same won’t happen with chat bot AIs? Bad actors (including PR agencies) will inevitably figure out where and how to spam comments in order to bias the AI models in favor of their agendas or products.
If the data they consume is filled with something like “fossil fuels don’t cause global warming because XYZ”, the chat bots will repeat it. They don’t have the capacity to reason.
There hasn’t been a reason to flood the internet with low effort spam because it’s easily detected by humans who read it. But the ML algorithms will be a lot easier to trick.
Injecting stuff into the data consumed by LLMs is the new type of SEO.
Apologies, I used the overly vague term “AI”. Any company creating an LLM that has web search + scraping capabilities will be at the mercy of the search results.
That said, LLMs are actually quite skilled at ignoring noise (repetitive data), so gaming SEO may lose popularity. Hell, the practice will DEFINITELY lose appeal once LLMs are just browsing for relevant content and summarizing without any citations (links to the sites). And even of they do cite, no one will click them.
Convenience > Fact
tldr; This additional layer of obfuscation between search and result will reshape the fabric of the internet with time
You can already outsource a lot of this to Bing. If you need to know the right temperature for making french fries, you can google a bunch of “recipes” (AKA life story of the author + history + vacation photos + cooking instructions) read them through and… actually better make some coffee while you’re at it because this is going to take a while. Anyway, the other option is to ask: “Hey Bing, I’m making french fries, but I don’t know how hot the oven should be.”
Spoiler: 220 °C
The scary thing is, what happens when people start doing this for more important things, such as what to do if your child has swallowed something or how to parallel park your car.
French fries aren’t made in an oven though.
Correct. However, if you buy frozen ones, you do need to heat them up some way. I ran out of nuclear weapons again, my flamer was out of gasoline, so using the oven was my best option.
Good thing those frozen ones come with the required cooking temperature on the package.
Hey Spez, can you throw some more subreddits into the dumpster fire. The temperature is almost right for popping some popcorn.
I think communicating with AI will become an art form the same way googling was/is.
Tea. Earl Grey. Hot
Until they’re sponsored
“I realize you seem frustrated from my responses. Nature’s Choice has a fantastic Stress Reducing gummy available at your local CVS”
Yeah, the gentle product hints at first will be driving people away quicker than a Monstered up Uber driver.
It already is. If you want to play a game of D&D with chatGPT, there’s a very specific prompt carefully crafted for that. If you want to chat with a with a total psycho, there’s a prompt for that. If you want your AI to do something it was specifically forbidden from doing, just craft a very specific prompt for that, and you’re good to go. You can even find sites that collect various prompts for just about any purpose you can imagine.
It’s the same idea I think, figuring out how to describe what you mean or phrase the question the right way to get the right kind of results.
Ask Jeeves was just ahead of its time