5 Embarrassingly Simple Things ChatGPT Can’t Do Yet.

1. ChatGPT Struggles to work with lists.

One embarrassing instance of this is the simple task of ordering a list chronologically.

If we construct a list like the one below:

1. Coinbase - 2012
2. TD Ameritrade - 1971
3. FTX - 2019.
4. PyTorch - 2016
5. Bloomberg Terminal - 1981
6. Charles Schwab - 1971
7. Robinhood - 2013
8. Interactive Brokers - 1977

??.

??..?

??… and so on.

ChatGPT will be unable to do something as simple as order the list chronologically by date. Our example below shows just how this occurs.

2. ChatGPT is overwhelmed by large prompts.

Depending on the request details, prompts involving more than 100 lines can cause ChatGPT to become overwhelmed. For example, if you hand ChatGPT a vertical list of entries just like the one we used above, and ask it anything… watch what happens.?

This is not only the case for lists. If we hand it a large enough paragraph it will also often become overwhelming and sometimes it will either completely forget the prompt accompanying the text. This often results in ChatGPT asking you if there is anything you would like it to help you do with your text, even if you already specified what you want from it before providing it with the overwhelmingly large context. And on other occasions, it just completely randomly jumps to answering a question you have never asked it before. This is especially obvious if you feed it an amount of data that most people would guess ChatGPT couldn't handle. In my research for example, whenever ChatGPT was provided with 5,000+ words it would always result in it becoming overwhelmed and responding to random unrequested prompts.

This last one may be embarrassing for both ChatGPT and users who have experienced it…?

3. ChatGPT Can't Handle Unlimited Requests

While TikTok would warn users that they’ve been scrolling the app for a ridiculous amount of time every so often, or how Netflix will ask if you’re even still watching if you binge for long enough, ChatGPT alerts users to their excessive usage based on request number within an hour.

We know unlimited may sound like a ridiculous standard to hold ChatGPT to, but consider your favorite video game or streaming service... it has never told you you are using it so much that it will now cut off your access. So while we don't expect ChatGPT to reach infinite computability, we do think it's a relatively beginner-level software problem to be unable to process as many requests as users are going to make with it.

4. Referencing prior requests or responses

(This one was discovered during the development of this article)

In addition to the limitations mentioned previously, ChatGPT also struggles to answer simple questions when they reference prior requests or responses. For example, if a user asks a follow-up question based on ChatGPT's previous response, it may not be able to answer it correctly. This is because ChatGPT does not have a memory like humans do and cannot keep track of previous requests or responses. As a result, it can sometimes give irrelevant or incorrect answers.

5. ChatGPT can be biased.

(This one was the only embarrassing incapability given to us directly by ChatGPT)?

Finally, ChatGPT has been shown to have biases, which can be embarrassing for both ChatGPT and its users. These biases can manifest in the form of gender, race, and other factors, and it is important to be aware of them when using ChatGPT.

While I can't say I've personally experienced these biases, during the editing of this article I used ChatGPT for feedback, and in one of its responses during the development of this article on its weaknesses, it unpromptedly added that one of its weaknesses was its biases. Maybe this hints at another weakness of its own, answering unprompted questions, or maybe it hints at its strength in that it will reveal information that users may find relevant even if it was not requested.

In conclusion, while ChatGPT is a remarkable achievement in the field of artificial intelligence, it still has its limitations. We must be aware of these limitations and use ChatGPT responsibly to avoid any embarrassing situations. However developers seem to be working hard on its improvement, and it is at least currently forthright enough to the extent that will reveal some of its weaknesses, even if it doesn't help ChatGPTs approval ratings. So with this, we can navigate its usage without taking everything it does as being the best way to go about doing it.