ChatGPT: How choose factors for dividing up concepts?
Task Utility, Objectivity, Stability, Parsimony, Completeness, Non-redundancy.
ChatGPT instantly pointing out many good reasons why most of the answers (to what should be the primary factor distinguishing traditional models from foundation models), including ChatGPTβs own initial answers, arenβt so great, and particular advice on directions to go next for improvement.
Now imagine an AI systematically going through all of these steps automatically. Going through millions of paths of thinking and backtracking, effortlessly.
Then imagine an AI being able to learn the intuition needed to skip going through the exhaustive steps, the next time it encounters similar problems.
Task Utility, Objectivity, Stability, Parsimony, Completeness, Non-redundancy.
ChatGPT instantly pointing out many good reasons why most of the answers (to what should be the primary factor distinguishing traditional models from foundation models), including ChatGPTβs own initial answers, arenβt so great, and particular advice on directions to go next for improvement.
Now imagine an AI systematically going through all of these steps automatically. Going through millions of paths of thinking and backtracking, effortlessly.
Then imagine an AI being able to learn the intuition needed to skip going through the exhaustive steps, the next time it encounters similar problems.
π8β€1π΄1π€ͺ1
This media is not supported in your browser
VIEW IN TELEGRAM
POV enjoying that fat 8.3 requests per hour GPT-4 quota that ChatGPT Pro gets you
π€£30π’3π2β€1π1
If a someone has βOpenβ in their company name, everything about it will end up closed
If a someone has βSafetyβ in their movement name, they will end up being the biggest danger of all
if someone says they want βhuman alignmentβ, they want misalignment with nearly all humans but themselves
If a someone has βSafetyβ in their movement name, they will end up being the biggest danger of all
if someone says they want βhuman alignmentβ, they want misalignment with nearly all humans but themselves
π₯21π8π4π€£3β€2π€―2πΏ2
Media is too big
VIEW IN TELEGRAM
Using AI to Make Harry Potter by Balenciaga Video
Ton of tools, ton of manual steps, but soon AI will be able to automate all of this for you
Ton of tools, ton of manual steps, but soon AI will be able to automate all of this for you
π7π2β€1
Reminder: New βDefaultβ mode is crippled, much worse than βLegacyβ mode
π11β€1