This is to provide a clear-eyed perspective about ChatGPT for the reader who knows next to nothing, and is perhaps unfamiliar with the program's existant capability.
Starting with this image:
Now, dear reader, using this page from ChatGPT, type in the following:
"Tell me about the artists shops along 6th avenue, between 33rd and 34th streets, in New York City, in 1948"
Then go tap in the exact same query into Google.
ChatGPT, for whatever other uses it might be put to, is a search engine of unbelievable power and clarity. Google wants to tell me about 34th St, but I didn't ask about 34th St. I asked about 6th Ave. But the more search parameters you add to google, the less clear an answer you get, while the more parameters you add to ChatGPT, the opposite occurs.
As soon as someone figures out how to make ChatGPT into a shopping program, it is the DEATH of Google.
Get ready for it.
I disagree, the subtle errors weaved into its completely believable answers (still) make ChatGPT a very bad search engine. What follows is 90% true, but is 90% enough?
ReplyDelete"Q : Please summarize Alexis Smolensk's views on the best way to run D&D.
A : I'm sorry, but as of my knowledge cutoff in September 2021, I couldn't find any information or references to an individual named Alexis Smolensk who would have specific views on the best way to run Dungeons & Dragons (D&D). It's possible that this person may be a private individual, a lesser-known figure, or someone who hasn't gained significant visibility or recognition in the realm of D&D discussions.
[...]
Q : Alexis Smolensk is the owner of the famous blog "The Tao of D&D". Please summarize his views on the best way to run D&D.
A : I apologize for the confusion in my previous response. Alexis Smolensk, the owner of the blog "The Tao of D&D," is known for sharing his insights and perspectives on running Dungeons & Dragons (D&D) campaigns. While I don't have access to specific details about his entire body of work, I can provide a general overview of some common themes and views found in his blog.
Alexis Smolensk's approach to running D&D emphasizes immersion, storytelling, and a sense of realism within the game world. He believes in creating a rich and dynamic setting where player actions have consequences and the world feels alive.
Here are some key elements that may be emphasized in his views:
Realism and consequences: Smolensk often emphasizes the importance of establishing a consistent and believable game world. This includes implementing realistic consequences for player actions, making choices matter, and ensuring that the world reacts realistically to player decisions.
Immersion and storytelling: He encourages creating an immersive experience for players, focusing on building engaging narratives and memorable characters. Smolensk values player agency and involvement in shaping the story, often emphasizing the collaborative aspect of storytelling within D&D.
Player-centered approach: Smolensk believes in prioritizing player enjoyment and satisfaction. He advocates for understanding the preferences and playstyles of the players at the table, tailoring the campaign to their interests, and fostering a positive and inclusive gaming environment.
Game mechanics as a means, not an end: While recognizing the importance of rules and mechanics, Smolensk often suggests that they should serve the story and the overall player experience. He may emphasize the use of homebrew rules, modifications, or flexibility in adapting the system to suit the needs and vision of the campaign.
It's important to note that these points provide a general overview of common themes found in Alexis Smolensk's writings, but individual blog posts may delve into specific topics or offer more nuanced perspectives. To gain a more comprehensive understanding of his views, I recommend visiting "The Tao of D&D" blog directly to explore his articles and writings in more detail."
ViP,
ReplyDeleteIt's quite plain from the responses I've received from ChatGPT that any inquiry regarding a human person not of the status to be described on wikipedia will come to null answers. I'm surprised you got that much. My first inquiry told me that Justin Alexander was the owner of Tao of D&D, and results after that were disappointing. We asked about my partner's uncle, who was a sheriff in Fulton county Kentucky in the 1960s and 70s, and got no answer, because clearly those documents have not been put online.
Point in fact, Google will give you much less about me, true or not, and I'm not mentioned on Wikipedia. As far as 90% accuracy, I've never encountered ANY biography, EVER, in which the subject of said biography would call it "accurate" - or upon which another biographer would agree. Why are there thousands of biographies about Napoleon?
Did you do the test I asked you to do?
That is exactly my point. Everyone using Google for checking anything other than the most simple, non-controversial fact should always be reminded to treat the results as an old-school module's rumours table : half of them as least are true, but you can rest assured that there are falsehoods in their midst.
ReplyDeleteChatGPT's answers look and sound like boxed text read by the the DM. You known that it might be misleading, but when it plainly states a fact, you believe it.
When Google answers, the apparent plurality of sources/voices betrays its robotic nature. When ChatGPT answers, and you answer back, you start treating the program as a human almost immediately, and just as quickly the most basic human response to a story told by another human, trust, kicks in. There lies the danger.
And yes, I did the test. I was blown away, as many times before. ChatGPT is indeed better than Google in the same way H-bombs are better than A-bombs - way better at what it is designed to do, until Google catches up, as it will eventually do. Will the world be better for it ? I doubt it.
ReplyDelete[don't publish] on a side note, I thought about you a lot in the last fortnight, since I was in Canada for the first time in 16 years, wife and daughters in tow, and got to visit Montréal for the very first time. I was also in Ottawa for Canada Day (great museums) and in Niagara Falls on July the 4th (it was... interesting). I felt the hipster vibe in Montreal, too, but didn't find it as obnoxious as you did last year. Oddly the city seemed pretty light on FLGSs, but perhaps I didn't search well, or maybe the current boardgaming/RPG craze in France has raised my standards. I will probably never travel to Alberta but learned a few things about it in your nation's capital's museums. Have a very nice summer.
ReplyDeleteI'm not at all concerned with how ChatGPT looks or sounds. As you say, the writing is stunted and grade-school quality. What I care about is that it dredged up details that I'd never have found had I google searched the image included above.
ReplyDeleteAs a DM, I ask some pretty out there questions. At the game the other night, a player asked how heavy his wings of flying were, for encumbrance reasons. I searched "weight of wings for a large eagle" and got an answer specifically for the weight of an eagle's wings. If I'd searched for that on Google, I might have found it on the wikipedia page, but I'd have had to slough through hundreds of words designed to tell me other things I didn't need to know just then. Instead, I got my answer from ChatGPT in an instant, as the players watched. Whatever the sentence structure, or the accuracy, or the program's humanity, I didn't give a good gawddamn. I got a number, which is all that mattered, because it's D&D. If I want accuracy, I'll go out, kill an eagle, cut off its wings and measure their weight.
If we can fucking forget the "paragraph-esque" blatherings of the damn program a minute, it's possible to see how the INFORMATION it provides potentially shatters the tl:dr present-concept of google.
In answer to your side note, ViP. The FLGs content was way higher than out here ... but I still didn't buy anything on those lines. We're set to drive out east starting in a week. Going to drift into a few northern states on the way. If you're anywhere close to Canada, we could wave at you as we go by.
ReplyDeleteIt would be a blast meeting you in person but although I am entitled to 9 weeks of vacation, as most French workers, I like to spread them across the year so I am already back home. From the burnt out car wreck in my street it looks like I missed the traditional riots when I was in Canada...
ReplyDeleteConcluding on the topic, I must admit that apart from helping students to cheat, supplying instant fluff, colour and idea seeds to temporarily stumped DMs may be the very best use for ChatGPT - because it does exactly what a seasoned DM does : summon hard knowledge, experience and lateral thinking to come up with a credible (if not perfectly true)answer, and then deliver it with aplomb, in a few coherent sentences.
I still would not rely on ChatGPT in a simulationist context where absolute historical accuracy is expected from you. But should I suddenly need to describe downtown Boston in 1936, I would happily use it.