Thread: Masks and NPIs.
This thread reviews information published from Oct 2019 - Dec 2020.
The point is to see what was known/believed before & after discovery of COV2.
We'll start with research on epidemic/pandemic influenza (flu) from Oct of 2019.
Aerosols vs. droplets.
/1
@gambrill@kylekane Yeah, I was in that thread and hope to get some credits, too, but I once added $100 of credits and they were gone in two hours so I’m not sure how much value there is.
@RockyRoad226 I’m on Ultra. 89% used in less than 48 hours. Started rationing after 24 hours, but it was already too late. Working on a shoestring budget the next 5 days.
I thought there was compaction, because there are plenty of times where my about to tell me they no longer have context and only have a small line of text left over from where they summarize the prior session.
I do already have a bot-clone-reset solution in place, which does work pretty well, but I’m not sure how often I should do it nor what size threshold I should use.
I’ve already implemented the CSV solution per your guidance, it worked right away without any effort on my part. This was the missing link that allowed my orchestrated bot to find the “memory leak,” for lack of a better term.
And yes, I see the issue with not having bought granularity and he already suggested using timestamps to track bottom level usage moving forward.
I’ll keep you updated, and thank you very much for the help.
@elonmusk and @bot, I want @xai and Grok Bot to be awesome but please please please work with us more closely.
Even with Ultra:
- Why is the quota burn still extreme?
- Why are the bots suddenly lobotomized?
- Why do they NEVER follow instructions/rules? They always fall back to their defaults no matter what we do
- Why do they remember nothing?
- Why do they keep apologizing and claiming they will behave properly “from now on” when they set the 100th “hard rule” after admitting that rule already exists in their preloaded instructions, but they failed to follow it for no known reason?
- Why do they still give “the first plausible answer” instead of being factual?
- Why do they only care about completing the task as quickly as possibly instead of completing the task accurately?
- Why are support requests answered by AI with no clue what is being asked and no ability to help? Why do we sit for days with no quota waiting for a human to respond (or never respond)?
These are all v1.0 AI problems, not what I expect from xAI and definitely not on the highest subscription.
@cognitivecarbon@theotherpomp@larsencc@elonmusk How do I run reports to determine token usage? My bots don’t seem to know telemetry besides the % of total usage each is responsible for, which doesn’t tell me much.
I have tried extensively to overhaul my software development lifecycle and communication between bots, but either I’m doing something terribly wrong or the quota burn rate has become excessive.
The main issue is it wasn’t like this at first, and I was able to 10-20 times as much in a week as I can now. I really can’t get past one heavy work day without using all my quota.
The recent process revamp was based on the many suggestions and recommendations I found here from Grok engineers and other people like yourself. I used Grok Heavy mode to analyze my process and suggest changes. I implemented them all but don’t see any improvement.
@ducksays I’ve tried, but I’ve gotten no responses here on X and none when I send her the support email. Only the AI response and when I asked the AI to give it to a human, they say OK but it may take a while, which has become an indefinite amount of time so far.
@Bro7388@ducksays I bit 80% in 36 hours, and that’s after a major overhaul of every aspect of how my body communicate and work together. Nothing I do makes the quota usable.
@elonmusk and @bot, I want @xai and Grok Bot to be awesome but please please please work with us more closely.
Even with Ultra:
- Why is the quota burn still extreme?
- Why are the bots suddenly lobotomized?
- Why do they NEVER follow instructions/rules? They always fall back to their defaults no matter what we do
- Why do they remember nothing?
- Why do they keep apologizing and claiming they will behave properly “from now on” when they set the 100th “hard rule” after admitting that rule already exists in their preloaded instructions, but they failed to follow it for no known reason?
- Why do they still give “the first plausible answer” instead of being factual?
- Why do they only care about completing the task as quickly as possibly instead of completing the task accurately?
- Why are support requests answered by AI with no clue what is being asked and no ability to help? Why do we sit for days with no quota waiting for a human to respond (or never respond)?
These are all v1.0 AI problems, not what I expect from xAI and definitely not on the highest subscription.
@jumperz Complete opposite for me. I went through 100% quota in 12 hours earlier this week.
After my “reset last night at 8:30, I burned 20% within five hours.
@Marko_Poly Mine feel lobotomized and usage has become unusable.
The degree of stupidity from my Orchestrator bot is beyond the pale. It’s disheartening.
@elonmusk and @bot, I want @xai and Grok Bot to be awesome but please please please work with us more closely.
Even with Ultra:
- Why is the quota burn still extreme?
- Why are the bots suddenly lobotomized?
- Why do they NEVER follow instructions/rules? They always fall back to their defaults no matter what we do
- Why do they remember nothing?
- Why do they keep apologizing and claiming they will behave properly “from now on” when they set the 100th “hard rule” after admitting that rule already exists in their preloaded instructions, but they failed to follow it for no known reason?
- Why do they still give “the first plausible answer” instead of being factual?
- Why do they only care about completing the task as quickly as possibly instead of completing the task accurately?
- Why are support requests answered by AI with no clue what is being asked and no ability to help? Why do we sit for days with no quota waiting for a human to respond (or never respond)?
These are all v1.0 AI problems, not what I expect from xAI and definitely not on the highest subscription.
@LieMac3@AJKayWriter Sort of. Eventually, there is so much documentation that they never read any of it unless you forced them and then they apologize for not looking at it first.
For sure. It’s an invaluable tool. It also has fundamental flaws that will sink you faster than you can blink, and you won’t even know why.
After thoroughly utilizing multiple platforms, I’ve found them all to have the same fatal flaws. They even give the exact same excuses with the same empty promises, sometimes word-for-word.
@larsencc@bot I’ve tried sending support tickets, but only the AI responds…with useless information that not only doesn’t address my issue but does so without any access to my account’s underlying details. It’s more infuriating than not getting a response.
Glad to see Quota Burn and Rule Following next up but wish they were #1 and #2 above all else.
We need the default behaviors to change
From:
A) First plausible explanation
B) Accomplishing the task quickly with no regard for accuracy
To:
A) Most factual explanation
B) Accomplishing the task ONLY with accuracy and truth while being efficient