4 ms·

In retrospect, such startups should have been wary: they should have known that OpenAI had Whisper, and also that GPT-4 was designed with image modality. I wou

by gwd 3y ago

In retrospect, such startups should have been wary: they should have known that OpenAI had Whisper, and also that GPT-4 was designed with image modality. I wouldn't say that OpenAI "telegraphed" their intentions, but the very first strategic question should have been, "Why isn't OpenAI doing this already, and what do we do if they decide to start?"

xeonmc 3y ago
Seems nobody learns from Sherlock.
Tenoke 3y ago
>I wouldn't say that OpenAI "telegraphed" their intentions They did telegraph it, they showed the multimodal capabilities back in the GPT4 Developer Livestream[0] right before first releasing it. 0. https://youtu.be/outcGtbnMuQ?t=943 https://youtu.be/outcGtbnMuQ?t=943
- makestuff 3y ago
  Yeah I remember watching that and thinking oh I know a cool app idea. What if you just take a video of what food is in your kitchen and Chat GPT will create a recipe for you. I go to the docs and that was literally the example they gave. I think the only place where plugins will make sense are for realtime things like booking travel or searching for sports/stock market/etc type information.
  - tornato7 3y ago
    I have a home-spun version of ChatGPT that uses function calling to connect to my emails, calendar, and notes. This is really useful because I can say "Bob just emailed me to set up a call. Respond to Bob with some available times from my calendar." That will be the real use case for plug ins.
shanusmagnus 3y ago
It would hard to be more explicit than doing a demo of multi-modality in GPT-4, and having an audio API that is amazing and that you can use right now, for pennies. It would be interesting to know if this really changed anything for anyone (competitors, VCs) for that reason. It's like the efficient market hypothesis applied to product roadmaps.
gzer0 3y ago
It is interesting that these startups did not recognize that the image modalities already existed, as evidenced by their initial GPT-4 announcement underneath “visual capabilities” [1]. [1] https://openai.com/research/gpt-4 https://openai.com/research/gpt-4

by gwd 3y ago

xeonmc 3y ago

Seems nobody learns from Sherlock.

Tenoke 3y ago

>I wouldn't say that OpenAI "telegraphed" their intentions They did telegraph it, they showed the multimodal capabilities back in the GPT4 Developer Livestream[0] right before first releasing it. 0. https://youtu.be/outcGtbnMuQ?t=943 https://youtu.be/outcGtbnMuQ?t=943

makestuff 3y ago
Yeah I remember watching that and thinking oh I know a cool app idea. What if you just take a video of what food is in your kitchen and Chat GPT will create a recipe for you. I go to the docs and that was literally the example they gave. I think the only place where plugins will make sense are for realtime things like booking travel or searching for sports/stock market/etc type information.
- tornato7 3y ago
  I have a home-spun version of ChatGPT that uses function calling to connect to my emails, calendar, and notes. This is really useful because I can say "Bob just emailed me to set up a call. Respond to Bob with some available times from my calendar." That will be the real use case for plug ins.

shanusmagnus 3y ago

It would hard to be more explicit than doing a demo of multi-modality in GPT-4, and having an audio API that is amazing and that you can use right now, for pennies. It would be interesting to know if this really changed anything for anyone (competitors, VCs) for that reason. It's like the efficient market hypothesis applied to product roadmaps.

gzer0 3y ago

It is interesting that these startups did not recognize that the image modalities already existed, as evidenced by their initial GPT-4 announcement underneath “visual capabilities” [1]. [1] https://openai.com/research/gpt-4 https://openai.com/research/gpt-4