New models already train on synthetic data. It's already a solved solution.
k110111
joined 1 year ago
Lol this reminds me of a time when I had KDE desktop environment installed on vanilla ubuntu. I thought I didn't really need ubuntu's default desktop environment and decided to 'purge' it. I quickly realized my f up when it deleted so many packages and ui started to act weird, I copied the shell's output to a file just incase, and sure enough I couldn't login with ui on next reboot. I was somehow able to login to shell and with some awk magic I was able to parse the text file to get all the packages I deleted and lo and behold everything worked just fine. Linux let's you f'up your OS but it also let's you fix it, it's just a skill issue.
I never buy any of the content but I do enjoy consuming content thus from a utilitarianistic point of view, piracy is a moral choice for me. Joking aside, if you can pay for things without a strong financial burden, you should pay.
Guys there are ways to get canva pro for free, check out fmhy.net or their Lemmy community.
I found this so far: https://github.com/KoljaB/RealtimeSTT
Maybe I can modify it to use whisper api.
Ligma 1 or ligma 2?
Except when it comes to defending their own country from their technologically inferior neighbor.
Its a disease, it has its own symptoms. If two people get a virus they will both catch a cold. If two people lost their job one of them might try to commit suicide while other one might only gets sad.
If you rearrange the letters in "sanders" it actually becomes hamas.
Guys the reason this study is important is because covid vaccines used revolutionary technology, they were the first to use mRNA based protein. If you remember we sequenced its genome within 40 days the making the vaccine was considerably easy. This is the main reason it took only 2 years for the vaccine to be made compared to years of development for other vaccines.
I used to have this now im mostly depressed
All the latest models are trained on synthetic data generated on got4. Even the newer versions of gpt4. Openai realized it too late and had to edit their license after Claude was launched. Human generated data could only get us so far, recent phi 3 models which managed to perform very very well for their respective size (3b parameters) can only achieve this feat because of synthetic data generated by AI.
I didn't read the paper you mentioned, but recent LLM have progressed a lot in not just benchmarks but also when evaluated by real humans.