21 ms·
Just had fun summarising the keynote using one of the new GPT-4 models (the 128K token model). It produced a pretty impressive blog post - not the kind of thin
by iamflimflam1 3y ago
Just had fun summarising the keynote using one of the new GPT-4 models (the 128K token model).
It produced a pretty impressive blog post - not the kind of thing I'd normally post, but certainly good enough for most "tech" publications... https://www.atomic14.com/2023/11/07/open-ai-dev-day.html https://www.atomic14.com/2023/11/07/open-ai-dev-day.html
- syntaxing 3y agoAha super meta. How did you transcribe the video? Whisper?
- jedberg 3y agoThe article says it used the transcript (since it's on YouTube I assume they mean the YouTube transcript).
- panarky 3y ago> not the kind of thing I'd normally post, but certainly good enough ... What would you change to make it better than "good enough"?
- michaelt 3y agoWell, the summary contains a lot of flowery, say-nothing language like "promising to reshape our world with unprecedented possibilities" and "groundbreaking developments that are set to further empower developers and users alike" where a human writing a summary would use less words. It retains video-only terms like "Join us as we explore" which would usually be omitted in a text summary. It defines the well-known acronym 'AI' then doesn't bother to define 'GPT' It has a section entitled "How GPTs Work" which doesn't say anything about how GPTs work. And it consistently calls the "Assistants API" the "Assistance API"
- coder543 3y agoWhat prompt did you use? (Besides the transcript itself)