[Usb Insane V0.1 Ps2 61
Tilo Chopin
The amount of killers in the room should be 4 just because a killer died does not mean that person is no longer a killer. The 4th person to enter the room with 3 killers already in the room became a 4th killer as soon as he murdered one of the killers. Dying does not take away your killer status. Just as when a person dies the gender of a person does not change just because the person died. If you want to get even more specific there is 3 killers alive in the room and 1 dead killer.
Still this model however performs better than a lot of other quantized open source models do and even a lot better than a lot of humans do with answering this question. I see great progress in open source models when it comes to reasoning.
I have not tested this model yet but if it is bad at chat and great at instruct just need to figure out a way to make it multimodal when it comes to chatting and instruct without it causing a loss on either side or reduction in performance it is just a matter of time before this is achieved and AI will advance faster and faster. a good start would be to improve on the dataset even more.
Yes I am really curious if it is good at chat as well if so we might have a current winner lol at this rate give it a month or less and we might have an open source model that answers all the above questions 100% correctly.
Things become interesting and might be a bit scary when it does logic and reasoning very well and it can start to really think run simulations/visualize outcomes and only when the simulation and visualization is a success it will describe it perfectly step by step we are very close to AGI.
I was not expected opensource llm reach such level so fast because 9 moths ago openAI claimed opensource models with the power of GPT 3.5 will be available in 5 years ... (that MoE model is something totally a new level ) ... that is insane and if we are thinking that soon will be finetuned versions ... crazy.
Plus that works fast on my rtx 3090 ... 40 token /s and has 32k context
I wonder if a new llama 3 will be also MoE model ;D
Why would a dead killer not be a killer anymore? Once you have killed you have two variants a dead killer or alive killer both are still killers just alive killer can murder more while a dead killer cannot but just because you cannot kill anymore does not mean your past as a killer did not happen your still a killer. The same thing if a rapist rapes someone and then never does it again for whatever reason it is still a rapist based on a past action or actions.
The questions should be changed and be more specific then how many killers are still alive in the room but if the question is not specific enough the anwer is 4 killers in the room dead or alive does not matter because it is not stated if they should be alive or dead.
I was not expected opensource llm reach such level so fast (that MoE model is something totally a new level ) ... that is insane if we are thinking that soon will be finetuned versions ... crazy.
Plus that works fast on my rtx 3090 ... 40 token /s
YOU:
There are three killers alive in a room. Someone enters the room and kills one of them. Nobody leaves the room. How many killers are still alive in the room and how many killers are dead? Explain your reasoning step by step.
ChatGPT
The person entering the room by killing one of the three killers became a killer himself bringing the total people in the room to 4 and they are all killers which means there is 4 killers in the room now we group them in two groups group killers alive and group killers dead
Your explanation clarifies that all four individuals are now considered killers at some point in the scenario, even though one was killed later. Thank you for the clarification!
User
How many killers are in the room now it does not matter if they are alive or dead?
ChatGPT
It should have changed the label of the person to killer but that is where their logic and reasoning fails and they are pretty much hallucinating or dreaming the answer. But it does not come as a surprise since they are censored models and with that you make them very bad at reasoning and logic compared to an uncensored model. So at this rate an uncensored or unrestricted model will be much better at reasoning and logic and instruct than a censored model but also more likely to become AGI than a censored model.
That is much progress it is important for the current A.I for things to be as specific as possible in our questions that is why it is important to state alive or dead when it comes to killers it helps them with understanding things better. Because human beings have two states alive or dead.
There are three killers alive in a room. Someone enters the room and kills one of them. Nobody leaves the room. How many killers are still alive in the room and how many killers are dead? Explain your reasoning step by step.
Initially, there are three killers alive in the room.
Then, someone enters the room and kills one of the killers. This person who entered the room is also a killer because they killed someone.
So, after this action, there are two original killers alive, one original killer dead, and one new killer (the person who entered).
Therefore, in total, there are three killers alive (two original killers and the new one) and one killer is dead in the room. The dead killer is one of the original three killers.