Gemanite55
Member
- Guildcard
- 0
Hi all!
Been a while since I posted in here, but I found something a little disturbing in the story I'm about to share, and thought this would be the place to share it.
"Ai" and LLMs have been around for a little while now, and I have a few benchmarks that I use to test and see whether or not they are good enough to be worried about them in any real capacity (beyond them displacing cubicle-jobs and stealing furry erotica.)
So far there's always been one reliable "benchmark" I use to trip them up. (Benchmark is in quotes because to me, a benchmark only matters if it's something I'd actually use, rather than actually benchmarking it's ability to replicate a novel or whatever.)
And that is just simply asking:
Most commonly it will come back with "I'm not allowed to look at specific websites," or, "I don't have access to the open internet," something like that.
The ones that get past that part are not much closer. Nearly all of them start with a decent first few steps up to level 10 or so then they usually add completely irrelevant information from other Phantasy Star games, or just seemingly any other game in it's language model. Some real "Neil Cicierega random word generator" stuff!
Well today, today my friends, this happened.
While using DeepSeek with DeepThink and Search enabled I gave it the same prompt from above, and this is what it gave me back:
(This is just the relevant information from the final section, the entire thing it gave back was much too long to post.)
I know that it's really not that special for a bot to essentially scrape info from this very forum and repost it, that's basically what Reddit has been for the last 10 years!
--and while there was a moment that I was super happy and very excited, "I don't have to use the tools, that the people from here have made, to create my own plans anymore!"
I realized: I don't have to use the tools that people made anymore.
For the last 30 minutes I've felt a bit empty and sad.
I keep envisioning a blacksmith looking up from forging a horseshoe and looking out at a world full of cars; then putting his hammer down with a pat of thanks.
Now I know that I don't HAVE to use an LLM to make my MAG feeding plans, just as much as I can choose to ride a horse places - or walk - and I know that LLMs are just some bubble that about to burst and whatever else - but that isn't the point.
Anyway, thanks for reading.
~JAX
Been a while since I posted in here, but I found something a little disturbing in the story I'm about to share, and thought this would be the place to share it.
"Ai" and LLMs have been around for a little while now, and I have a few benchmarks that I use to test and see whether or not they are good enough to be worried about them in any real capacity (beyond them displacing cubicle-jobs and stealing furry erotica.)
So far there's always been one reliable "benchmark" I use to trip them up. (Benchmark is in quotes because to me, a benchmark only matters if it's something I'd actually use, rather than actually benchmarking it's ability to replicate a novel or whatever.)
And that is just simply asking:
I have tried this with the free version of just about any of the available LLMs over the last 2-ish years, and so far none of them have even been close!
Most commonly it will come back with "I'm not allowed to look at specific websites," or, "I don't have access to the open internet," something like that.
The ones that get past that part are not much closer. Nearly all of them start with a decent first few steps up to level 10 or so then they usually add completely irrelevant information from other Phantasy Star games, or just seemingly any other game in it's language model. Some real "Neil Cicierega random word generator" stuff!
Well today, today my friends, this happened.
While using DeepSeek with DeepThink and Search enabled I gave it the same prompt from above, and this is what it gave me back:
###Step-by-Step Feeding Plan:
#### Phase 1: Levels 1–35 (Ranger with Any Section ID)
- **Goal**: Obtain Estlla (Lv.10) and Pilla (Lv.35) while maximizing DEX.
- **Items**: Feed **Antidotes** exclusively (from Table 1: +15 DEX, –0 POW, –0 MIND per feed; later evolutions may use Table 4 or 7 for higher gains) .
- **Evolution**:
- **Lv.10**: Evolves to **Kalki** (Ranger) → Gains **Estlla** .
- **Lv.35**: Evolves to **Mitra** (DEX highest) → Gains **Pilla** .
- **Stats at Lv.35**: Approximately `5/7/23/0` (varies slightly due to fractional gains).
#### Phase 2: Levels 36–49 (Continue on Ranger)
- **Goal**: Prepare for Mylla & Youlla at Lv.50.
- **Items**: Continue feeding **Antidotes** to maintain DEX as the highest stat.
- **Stats at Lv.49**: Approximately `5/7/37/0`.
- **Transfer**: At Lv.49, move mag to a **Hunter or Ranger with an even Section ID** (e.g., Skyly, Purplenum) .
#### Phase 3: Level 50 (Hunter/Ranger with Even Section ID)
- **Goal**: Obtain Mylla & Youlla.
- **Items**: Feed one **Antidote** to reach Lv.50.
- **Evolution**: Evolves to **Ila** (Hunter) or **Kaitabha** (Ranger) → Gains **Mylla & Youlla** .
- **Stats at Lv.50**: Approximately `5/7/38/0`.
#### Phase 4: Levels 51–100 (Achieve Evolution Stats: 5/45/50/0)
- **Goal**: Reach stats `5/45/50/0` by Lv.100 to satisfy `DEF + POW = DEX + MIND` for Nidra evolution.
- **Strategy**: Use a combination of **Dimates** (Table 7: +27 POW, –10 DEX, –16 MIND) and **Antidotes** (Table 7: +28 DEX, –10 POW, –10 MIND) to net-increase POW and DEX without affecting DEF or MIND .
- **Calculated Feeds**:
- Increase POW from ~7 to 45 (+38 levels): Feed ~43 Dimates (net +27 POW each after compensating with Antidotes).
- Increase DEX from ~38 to 50 (+12 levels): Feed ~13 Antidotes (net +28 DEX each).
- **Banking**: Deposit/withdraw mag every 10–20 feeds to reset the timer and manage fractional stats (e.g., after every 10 feeds to optimize gains) .
- **Transfer**: At Lv.99, move mag to a **Male Force** with required Section ID (Viridia, Bluefull, Redria, or Whitill) .
- **Lv.100**: Feed one item (e.g., Antidote) to evolve to **Nidra**.
#### Phase 5: Levels 101–200 (Final Stats: 5/133/62/0)
- **Goal**: Increase POW from 45 to 133 (+88 levels) and DEX from 50 to 62 (+12 levels).
- **Strategy**: Continue using **Dimates** and **Antidotes**:
- **Dimates**: Feed ~394 Dimates (net +27 POW each) to gain +88 POW.
- **Antidotes**: Feed ~184 Antidotes (net +28 DEX each) to gain +12 DEX.
- **Banking**: Bank periodically (e.g., every 20 feeds) to optimize feeding efficiency and manage fractional gains .
(This is just the relevant information from the final section, the entire thing it gave back was much too long to post.)
I know that it's really not that special for a bot to essentially scrape info from this very forum and repost it, that's basically what Reddit has been for the last 10 years!
--and while there was a moment that I was super happy and very excited, "I don't have to use the tools, that the people from here have made, to create my own plans anymore!"
I realized: I don't have to use the tools that people made anymore.
For the last 30 minutes I've felt a bit empty and sad.
I keep envisioning a blacksmith looking up from forging a horseshoe and looking out at a world full of cars; then putting his hammer down with a pat of thanks.
Now I know that I don't HAVE to use an LLM to make my MAG feeding plans, just as much as I can choose to ride a horse places - or walk - and I know that LLMs are just some bubble that about to burst and whatever else - but that isn't the point.
Anyway, thanks for reading.
~JAX
