Thoughts about AI vs Forums

[jsteele]

As the owner of a discussion forum, I obviously am self-interested in the success of discussion forums, particularly this one. One of the prime benefits of forum such as this is to get answers to questions. In the simplest case, a user comes to DCUM, posts a question, and gets answers. With many questions, there is no "correct" answer. Rather, there can be several answers, all of which might be correct with various caveats. For instance, if someone asked which neighborhood in DC is the best in which to live in order to attend a good school, there could be various "correct" answers. For instance, someone might say AU Park because it is zoned for Janey, Deal, and Jackson-Reed. Another answer might be Petworth because it is convenient to a number of popular Public Charter Schools. A number of neighborhoods, both east and west of Rock Creek Park, might be named due to their easy commutes to Sidwell, GDS, the Cathedral schools, and a number of other popular schools. Posters are able to debate the pros and cons of each of these answers and the original poster can arrive at their own informed conclusion.

For years, posters have sought answers to their questions by using a search engine, particularly Google. "Googling" something didn't mean that Google would provide the answer to the question. Rather, Google linked to resources — sometimes even DCUM — that could provide the answer. Over the years Google has probably been our most significant source of traffic.

Now, with the spread of AI, search engines are providing AI-generated answers to questions. If you ask Google, "what neighborhood in DC is best for good schools", Google will provide an AI-generated response. This response is based on data that Google has used to train its AI system. That data has not been vetted for accuracy. When I tested this with Google, it recommended Cleveland Park, Georgetown, and Woodley Park. With regard to Cleveland Park, it mentioned Washington International School and Eaton Elementary. Georgetown was suggested because of School Without Walls and Washington Latin Public Charter School. There was no explanation about why Woodley Park was suggested other than it is good for families looking for schools. I am sure that many of us can find reasons to object to this response. This is the problem with AI, at least in its current state. Users are led to believe that the answers are correct when there is a good chance they are not. There is no option for an alternative view. Certainly, Google's AI-generated response would provide responses criticizing it if the answer were posted on DCUM.

While that is but one example, it is fairly representative of the state of AI and search engines. Unvetted data used to train AI models may be correct or it may not be. Users have no way of knowing. Due to this shortcoming, I think that users, at least smart ones, will lose trust in the AI-generated responses and begin to rely on forums in which responses come from real people. Those responses may well be wrong, but there is a good chance that another human will point out the errors. Moreover, when posting on a forum, most users know to be discerning about responses and not immediately trust that replies will be accurate. Search engines train users to accept their responses without question.

The bottom line of my argument in this post is that the medium that many thought was old fashioned and would soon fall into disfavor if not disuse, is likely to be the the real wave of the future. While AI models linked to search engines become oracles of undependable "facts", discussion forums will be market places of ideas in which crowd-sourced responses have built in quality control provided by other users.

[Anonymous]

Sounds about right. AI is close to peak hype just now.

[Anonymous]

I don’t know people think AI is a fad that will go away. It is here to stay and is only going to improve. My guess is they think that because they don’t understand it.

I think you will lose some traffic due to AI answers. That’s a common concern for site owners. I use the Brave browser a fair bit and it produces AI responses to many things I search for. I make do with those sometimes, but I will still click on search hits fairly often. It has definitely reduced my viewing of sites somewhat though.

My guess is there will still be a lot of activity from people looking for local intel, feedback on health issues and all that. Some people need a place to vent too. Plus it’s no fun arguing politics with AI.

[Anonymous]

Forum posts are guided by anecdata, biases, emotion, "Googling", and trolling for that matter. AIs don't simply regurgitate what they've been trained on. What forums provide is the chance to engage with humans, not chatbots.

[Anonymous]

You raise an interesting point but what I’ve seen which drives me crazy is people responding to posts by pasting in AI responses!

[Anonymous]

I am pretty ignorant on how AI collects its data, but I imagine that it is less useful when answering questions about current trends.
For example, there are a lot of questions in the real estate forum about payment structures for real estate agents now that the buyer's agent is no longer paid by the seller. I am selling a home now so I see that agents are doing it different ways and some seller agents are still asking their clients to pay the buyers agent. So it is in flux until the industry settles into a new pattern.
I imagine that AI would give an outdated answer. Now to be fair, a lot of answers in that forum are speculation from people who have no idea of what the market is currently doing but those answers are not necessary wrong.

[Anonymous]

One issue for anonymous forums like DCUM is going to be the influx of "helpful" users who are trying to be helpful and provide unfiltered AI generated answers.

The reason I use, and contribute to, forums is the community. Places like Stack Exchange and Reddit try and gamify the community. Here, the community comes about from some of the inside "jokes" (e.g., the money threads about 1M+ HHI being UMC) and some of the personal anecdotes that AI is not yet trying to mimic.

If I want numbers/facts about a neighborhood (school ranking, diversity, etc) Google and AI are reasonable places to start. If I want personal insight, I am going to ask here.

[Anonymous]

I think people use Google and specifically AI for much different reasons than going to an Internet forum. I Google something if I just want a straight answer. I go to a forum because I want to be a part of a discussion. I don’t think Google/AI is a threat to forums.

[Anonymous]

Anonymous wrote:I don’t know why people think AI is a fad that will go away. It is here to stay and is only going to improve. My guess is they think that because they don’t understand it.

We have watched at least one boom/bust AI hype cycle per decade since the 1980s. Each time legitimate progress towards AI has happened, each time more AI gets used, but each time it was much less progress than the hype suggested. And each bust evidenced a surplus of recent grads who focused in on AI - meaning disproportionately large layoffs among the AI crowd.

The current hype cycle is driven by LLMs and is a legitimate improvement, but people are starting to realize LLMs hallucinate and are helpful for a much smaller set of problems than the hype suggests.

[jsteele]

Anonymous wrote:You raise an interesting point but what I’ve seen which drives me crazy is people responding to posts by pasting in AI responses!

This bugs me as well and I delete such posts when I see them.

[Anonymous]

Anonymous wrote:I am pretty ignorant on how AI collects its data, but I imagine that it is less useful when answering questions about current trends.
For example, there are a lot of questions in the real estate forum about payment structures for real estate agents now that the buyer's agent is no longer paid by the seller. I am selling a home now so I see that agents are doing it different ways and some seller agents are still asking their clients to pay the buyers agent. So it is in flux until the industry settles into a new pattern.
I imagine that AI would give an outdated answer. Now to be fair, a lot of answers in that forum are speculation from people who have no idea of what the market is currently doing but those answers are not necessary wrong.

No, ChatGPT (and likely others) no longer have this limitation. The LLM accesses current data as part of its response construction (the LLM is simply the language component of that processing). Go ahead and ask it what the current trends are. It'll say that it's currently in flux but options right now are direct payment by the buyer, flat fees or hourly rates, etc. It actually provides a more coherent response than I was able to find on the internet.

[Anonymous]

Anonymous wrote:

Anonymous wrote:I don’t know why people think AI is a fad that will go away. It is here to stay and is only going to improve. My guess is they think that because they don’t understand it.

We have watched at least one boom/bust AI hype cycle per decade since the 1980s. Each time legitimate progress towards AI has happened, each time more AI gets used, but each time it was much less progress than the hype suggested. And each bust evidenced a surplus of recent grads who focused in on AI - meaning disproportionately large layoffs among the AI crowd.

The current hype cycle is driven by LLMs and is a legitimate improvement, but people are starting to realize LLMs hallucinate and are helpful for a much smaller set of problems than the hype suggests.

Technology wasn’t as it is today, and neither was the internet. AI is being rapidly adopted into customer service, phones and many other products. The amount of money and research being done will produce substantial advances.

[Anonymous]

This forum is a fundamentally different form of social connection and infotainment than user interactions with ai tools.

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:I don’t know why people think AI is a fad that will go away. It is here to stay and is only going to improve. My guess is they think that because they don’t understand it.

We have watched at least one boom/bust AI hype cycle per decade since the 1980s. Each time legitimate progress towards AI has happened, each time more AI gets used, but each time it was much less progress than the hype suggested. And each bust evidenced a surplus of recent grads who focused in on AI - meaning disproportionately large layoffs among the AI crowd.

The current hype cycle is driven by LLMs and is a legitimate improvement, but people are starting to realize LLMs hallucinate and are helpful for a much smaller set of problems than the hype suggests.

Technology wasn’t as it is today, and neither was the internet. AI is being rapidly adopted into customer service, phones and many other products. The amount of money and research being done will produce substantial advances.

That is roughly what was said in the 1990s, in the early 2000s, and in the 2010s. Time will tell. Based on experience, I remain skeptical.

Other poster also is correct that “customer service” and similar uses you mention are irrelevant to forums like DCUM.

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:I don’t know why people think AI is a fad that will go away. It is here to stay and is only going to improve. My guess is they think that because they don’t understand it.

We have watched at least one boom/bust AI hype cycle per decade since the 1980s. Each time legitimate progress towards AI has happened, each time more AI gets used, but each time it was much less progress than the hype suggested. And each bust evidenced a surplus of recent grads who focused in on AI - meaning disproportionately large layoffs among the AI crowd.

The current hype cycle is driven by LLMs and is a legitimate improvement, but people are starting to realize LLMs hallucinate and are helpful for a much smaller set of problems than the hype suggests.

Technology wasn’t as it is today, and neither was the internet. AI is being rapidly adopted into customer service, phones and many other products. The amount of money and research being done will produce substantial advances.

That is roughly what was said in the 1990s, in the early 2000s, and in the 2010s. Time will tell. Based on experience, I remain skeptical.

Other poster also is correct that “customer service” and similar uses you mention are irrelevant to forums like DCUM.

Its use in customer service applications is about as close as you can get to the discussion around using message boards as an intellectual resource.