Call it AI if you want but it's simple theft, same old google since day one . Danny's example here is a perfect example, "AI" knows nothing about Bryce Canyon or Arches National Park without the work of actual people.
I totally agree with you... I would like to know how will Google compensate the website owners - those that actually create content.
Shepherd
10:56 am on May 17, 2023 (gmt 0)
I, for one, welcome our new AI overlords...
Listen AI, that's cute and all with the satisfying search results. I imagine I might have a more satisfying car buying experience if the car dealership took the car from the manufacture and gave it to me for free.
superclown2
11:31 am on May 17, 2023 (gmt 0)
It seems that legislators are keen on regulating AI. I'm not sure how they can since even those who created it don't seem to know what it's going to do. No doubt Europe will tame it years before the USA people finish lapping up the 'lobbying' cash.
Satisfying search results? Just show me what I asked for and not what you think I should see, and what is the most profitable for you. If I type in pink left handed unicorns show me sites that actually feature them, and not the pet shop that pays the most for ads. Simple, and no need for a mega expensive artificial intelligence that sometimes wonders off into fairyland.
TechNoob
4:24 pm on May 23, 2023 (gmt 0)
I've been hearing declarations of SEO "dying" because of all these AI innovations. I don't think it is time for us to hang up our hats yet, but I have to sleep with one eye open, right?
Shepherd
11:47 am on May 24, 2023 (gmt 0)
Assuming attribution requirements get put in place (ChatGPT seems to be giving proper attribution from what I've seen so far) we might be talking about a change instead of a death, SEO -- AIO
engine
12:32 pm on May 24, 2023 (gmt 0)
SEO adapts as long as there are search engines. When AdWords was launched it spurned a whole new sector. There's no reason to assume this will not be the case again. I'm staying positive.
tangor
12:47 am on May 27, 2023 (gmt 0)
I am positive that "AI for me, but not for thee" is going to be mainstream from here on out!
FranticFish
5:40 am on Jun 2, 2023 (gmt 0)
In terms of theft, Large Language Model answers are not really as bad as a greedy search engine in some ways. Answer boxes are directly lifted from someone's site. The major LLMs are given staggering amounts of data * so their responses are far less direct theft.
* I gather that the cost of the storage required to seed the word association probability tricks used to generate responses is a major (or maybe even THE major) expense.
In terms of competition, it's possible that LLMs might offer a better opportunity to take aim at the big boys and fight back.
Before LLMs, Google increasingly shut content providers out of monetising their sites, and there was no real way for groups of smaller businesses to band together and offer a viable alternative. Whatever bias there was in the way information was presented and blended with ads, the platform itself and the way everything was integrated was fantastic.
I've only just started to try to educate myself about this. Would welcome corrections if I've got anything wrong.
Microsoft & Google may have dropped billions on their LLMs (and as I said, I think much of this is storage cost), but there are Open Source ones that you can run on a laptop, and someone even got one running on a phone. People have trained LLMs in a matter of days on niche topics.
If a collective of local or niche businesses decided to come together to create and train their own topic-specific LLM and then run it via an app and/or hosted online, they might be able to offer something better than the titans who are trying to be all things to everyone in two ways:
1) Greater accuracy. An LLM trained on carefully curated data by people with specialist subject matter knowledge would be likely to be more accurate and 'hallucinate' less.
2) More up to date. GPT for example was initially only trained up to 2021. I gather it does now 'know' more recent things (e.g. that Elon Musk is now CEO of Twitter) but I don't think it has been trained on ALL new knowledge since 2021 because it is very labour-intensive to do this. LLMs work on the statistical relationships between words in a corpus of content, and I don't know whether it is or is not possible to only train them partially (i.e. 'top them up'). Whichever it is though, the less information you are trying to cover and to keep relevant/current, the easier the job becomes.
tangor
8:41 am on Jun 3, 2023 (gmt 0)
Keeping LLMs up-to-date (AND ACCURATE) is going to be both time consuming and expensive for anything that is not general knowledge. LATEST news/fads are actually too new for LLM since the sample size will be so much smaller than "older" information, and accuracy will be hampered equally since there will not be enough corroboration on that same data (too new).
Jury is still out on the value of AI... not enough history/proof it actually does what it claims. THAT SAID, AI is here to stay and we'll have to see how it all plays out.
