4 ms·

> The Star Trek computer from TNG is basically an LLM, really. The Star Trek computer is not like LLMs: a) it provides reliable answers, b) it is capable of re

> The Star Trek computer from TNG is basically an LLM, really. The Star Trek computer is not like LLMs: a) it provides reliable answers, b) it is capable of reasoning, c) it is capable of actually interacting with its environment in a rational manner, d) it is infallible unless someone messes with it. Each one of these points is far in the future of LLMs.

sgt 1y ago
Yet when you ask it to dim the lights, it dims either way too little or way too much. Poor Geordi.
- sgt 1y ago
  For what it's worth, I was referring to the episode when he set up a romantic dinner for the scientist lady. Computer couldn't get the lighting right.
lcnPylGDnU4H9OF 1y ago
Their point is that it seems to function like an LLM even if it's more advanced. The points raised in this comment don't refute that, per the assertion that each of them is in the future of LLMs.
- NoTeslaThrow 1y ago
  > Their point is that it seems to function like an LLM even if it's more advanced. So did ELIZA. So did SmarterChild. Chatbots are not exactly a new technology. LLMs are at best a new cog in that same old functionality—but nothing has fundamentally made them more reliable or useful. The last 90% of any chatbot will involve heavy usage of heuristics with both approaches. The main difference is some of the heuristics are (hopefully) moved into training.
  - Philpax 1y ago
    Stating that LLMs are not more reliable or useful than ELIZA or SmarterChild is so incredibly off-base I have to wonder if you've ever actually used a LLM. Please run the same query past ELIZA and Gemini 2.5 (https://aistudio.google.com/prompts/new_chat https://aistudio.google.com/prompts/new_chat) and report back.
    - NoTeslaThrow 1y ago
      > Please run the same query past ELIZA and Gemini 2.5 (https://aistudio.google.com/prompts/new_chat https://aistudio.google.com/prompts/new_chat) and report back. I don't see much difference—you still have to take any output skeptically. I can't claim to have ever used gemini, but last I checked it still can't cite sources, which would at least assist with validation. I'm just saying this didn't introduce any fundamentally new capabilities—we've always been able to GIGO-excuse all chatbots. The "soft" applications of LLMs have always been approximated by heuristics (e.g. generation of content of unknown use or quality). Even the summarization tech LLMs seem to offer don't seem to substantially improve over the NLP-heuristic-driven predecessors. But yea, if you really want to generate content of unknown quality, this is a massive leap. I just don't see this as very interesting.
      - filoleg 1y ago
        > I can't claim to have ever used gemini, but last I checked it still can't cite sources, which would at least assist with validation. Yes, it can cite sources, just like any other major LLM service out there. Gemini, Claude, Deepseek, and ChatGPT are the ones I personally validated this with, but I bet other major LLM services can do so as well. Just tested this using Gemini with “Is fluoride good for teeth? Cite sources for any of the claims” prompt, and it listed every claim as a bullet point accompanied by the corresponding source. The sources were links to specific pages addressing the claims from CDC, Cleveland Clinic, John Hopkins, and NIDCR. I clicked on each of the links to verify that they were corroborating what Gemini response was saying, and they were. In fact, it would more often than not include sources even without me explicitly asking for sources.
        1y ago

sgt 1y ago

Yet when you ask it to dim the lights, it dims either way too little or way too much. Poor Geordi.

sgt 1y ago
For what it's worth, I was referring to the episode when he set up a romantic dinner for the scientist lady. Computer couldn't get the lighting right.

lcnPylGDnU4H9OF 1y ago

Their point is that it seems to function like an LLM even if it's more advanced. The points raised in this comment don't refute that, per the assertion that each of them is in the future of LLMs.

NoTeslaThrow 1y ago
> Their point is that it seems to function like an LLM even if it's more advanced. So did ELIZA. So did SmarterChild. Chatbots are not exactly a new technology. LLMs are at best a new cog in that same old functionality—but nothing has fundamentally made them more reliable or useful. The last 90% of any chatbot will involve heavy usage of heuristics with both approaches. The main difference is some of the heuristics are (hopefully) moved into training.
- Philpax 1y ago
  Stating that LLMs are not more reliable or useful than ELIZA or SmarterChild is so incredibly off-base I have to wonder if you've ever actually used a LLM. Please run the same query past ELIZA and Gemini 2.5 (https://aistudio.google.com/prompts/new_chat https://aistudio.google.com/prompts/new_chat) and report back.
  - NoTeslaThrow 1y ago
    > Please run the same query past ELIZA and Gemini 2.5 (https://aistudio.google.com/prompts/new_chat https://aistudio.google.com/prompts/new_chat) and report back. I don't see much difference—you still have to take any output skeptically. I can't claim to have ever used gemini, but last I checked it still can't cite sources, which would at least assist with validation. I'm just saying this didn't introduce any fundamentally new capabilities—we've always been able to GIGO-excuse all chatbots. The "soft" applications of LLMs have always been approximated by heuristics (e.g. generation of content of unknown use or quality). Even the summarization tech LLMs seem to offer don't seem to substantially improve over the NLP-heuristic-driven predecessors. But yea, if you really want to generate content of unknown quality, this is a massive leap. I just don't see this as very interesting.
    - filoleg 1y ago
      > I can't claim to have ever used gemini, but last I checked it still can't cite sources, which would at least assist with validation. Yes, it can cite sources, just like any other major LLM service out there. Gemini, Claude, Deepseek, and ChatGPT are the ones I personally validated this with, but I bet other major LLM services can do so as well. Just tested this using Gemini with “Is fluoride good for teeth? Cite sources for any of the claims” prompt, and it listed every claim as a bullet point accompanied by the corresponding source. The sources were links to specific pages addressing the claims from CDC, Cleveland Clinic, John Hopkins, and NIDCR. I clicked on each of the links to verify that they were corroborating what Gemini response was saying, and they were. In fact, it would more often than not include sources even without me explicitly asking for sources.
      - 1y ago