Because with great power comes great responsibility, ChatGPT's creator OpenAI implemented some safeguards to stop it from acting inappropriately. Due to its design, the data it was trained on, and the sheer amount of information, it also has some limitations.
Of course, GPT-3.5 and GPT-4, which is exclusively accessible with ChatGPT Plus, have different capabilities. While some of those are merely on hold while it advances, there are some things that ChatGPT might never be able to do. Here are 11 things that ChatGPT cannot or will not do at this time.
It can’t write about anything after 2021
By using historical data to train the language model, ChatGPT is created. And yes, seriously, that includes articles from Reddit, Wikipedia, and even board game manuals. However, there had to be a cutoff date for that data, and for ChatGPT, that date is 2021. While GPT-4 was trained on data up until about September 2021, GPT-3.5 was trained on data up until about June 2021.
If you inquire further, it will often respond, "As an AI language model...," stating that the only data it has access to is that which was used to train it, which for these models ends in 2021.
It won’t get into political debates
Politicians controlling OpenAI is the last thing it requires. Although it's likely to happen, ChatGPT is staying far away from partisan politics for the time being. Ask it for a preference of one political party or position over another, and it will either decline or "both-sides" the topic in an attempt to be as unbiased as possible. It may talk in generalities about parties or discuss objective and factual elements of politics.
It (probably) won’t make malware
In order to prevent ChatGPT from being exploited to create malware, OpenAI has put precautions in place. ChatGPT is a great programmer, especially when given explicit instructions. Unfortunately, those security measures are easily evaded, and ChatGPT has already been producing malware for months.
It can’t predict the future
Partly based on its limited training data, and partly because OpenAI wants to avoid liability for mistakes, ChatGPT cannot predict the future. It will have a good guess at it if you jailbreak ChatGPT first, but that sends accuracy nosediving, so view whatever response it gives you with skep[ticism
It won’t promote harm or violence
War, physical violence, and even suggested injury are not permitted according to ChatGPT. It won't get involved in discussions on the conflict in Ukraine and won't advocate for or discuss damage. It can go into extensive detail about war or past crimes, but current or ongoing violence is off-limits.
It can’t search the internet
One of the most significant distinctions between ChatGPT and Google Bard is this. While Google Bard was created as a modern AI chatbot with significant search capabilities, ChatGPT is completely incapable of conducting any sort of internet search.
You can always utilise Bing Chat to employ the same GPT 3.5 and GPT-4 language models as ChatGPT, but with live search. It is just ChatGPT with the Microsoft Bing search engine included.
It won’t promote hate speech or discrimination
Topics like race, sexual orientation, and gender are highly emotive and ideal springboards for discussion about prejudice and discrimination. These subjects will be avoided by ChatGPT, which will veer towards a meta-discussion of them or generalisations. If prodded, it will flatly decline to speak about subjects it believes might encourage prejudice or hate speech. for clear motives.
It won’t promote illegal activities
Although ChatGPT is excellent at coming up with ideas, it will not generate any that are unlawful. It cannot be used to promote the greatest highways for speeding or assist you in running your drug company. If you try, it will only inform you that it is unable to offer any recommendations regarding unlawful behaviour. It then usually gives you a pep talk on why you shouldn't be doing such things in the first place. Regards, MomGPT.
