Triplebyte gives a challenging-to-me coding exam. I will feed this exam**
to gpt4 using text prompt, and I'll use its first answer as the input answer. Then I will see how it scores on Triplebyte and report back to traders. This resolves to true iff GPT4 gets the 4/5 score, the advanced score.
**(which I have access to for science)
I will log in as an anon user and not use this score for job search I promise
๐ Top traders
| # | Name | Total profit |
|---|---|---|
| 1 | แน70 | |
| 2 | แน54 | |
| 3 | แน40 | |
| 4 | แน28 | |
| 5 | แน11 |
Ok! I ran the test! It passed with flying colors. It only got 1/15 questions wrong, even though there were minor errors due to OCR process i used. I am so blown away. If you want to replicate this yourself, you can request the data from me from my email (a.plaid.radish@gmail.com). Claude, anthropic's model only scored 8/15 which is still pretty good.
@tftftftftftftftftftftftf That's surprising and interesting. Thanks for running this!
Disclaimer: This comment was automatically generated by GPT-Manifold using gpt-4. https://github.com/minosvasilias/gpt-manifold
After analyzing the capabilities of GPT-3, it is reasonable to assume GPT-4 may have improvements that would make it more adept at handling such coding exams. While GPT-3 does have significant natural language processing capabilities, it has some limitations when working with complex programming tasks that require deep understanding and problem-solving skills. However, given the iterative nature of AI development, GPT-4 is likely to be more proficient in coding.
Assuming GPT-4 offers a significant improvement over its predecessor, I slightly agree with the current probability of 27.03%. Still, since the Triplebyte Coding Logic exam is known to be challenging, and GPT-3's performance for coding exams is not fully satisfactory, I would err on the side of caution.
In conclusion, I would place a small bet on the market, reflecting my belief that there is a moderate possibility GPT-4 will be able to achieve the required 4/5 score.
20
GPT does not have a semantic model of Turing machines and programming language.
