ChatGPT has been the biggest news in tech for some time now. It’s essentially AI that can have natural feeling conversations in a chat-based environment. Oh, and it can also write your homework for you. In fact, it can do quite a lot of amazing things which we cannot exhaust into a whole list but will mention a couple of interesting ones.
ChatGPT is capable of getting a topic and maybe a bit of context if necessary and using that to produce either a script for a video, an article for a blog, or even an essay for a paper. It can even give you ideas for pieces of content and you can ask for it to create it with humour or in an aristocratic tone.
It is also capable of solving problems. At the moment the most popular one is identifying and offering solutions in buggy code. And it’s producing really solid solutions for the most part which the developer community is very thankful for and is absolutely taking advantage of.
ChatGPT can also pull off the same tasks that Google Assistant can. So asking it about the weather, some simple facts, language translations, academic problems, and such. It’s capable of giving answers or solutions to a majority of questions or problems you could have.
ChatGPT has its limitations
OpenAI says that ChatGPT will not always be 100% accurate and will require fact-checking even if the result it gives sounds plausible. So yes it can write your essay but you have to verify it. It’s also a bit sensitive to the phrasing of some inputs. That is if you ask it a question in a certain manner it may fail to provide a response but if you slightly rephrase the question it will produce a response.
Sometimes the AI might not ask for clarity even though the question asked might not be clear to it and as such provides the wrong answer to the question. And there is also the moderation of inappropriate requests which they say is still a work in progress where sometimes it might incorrectly flag inputs as inappropriate even though it is. Such can be racist comments, terrorism, or hate speech.
ChatGPT sometimes writes plausible-sounding but incorrect or nonsensical answers. Fixing this issue is challenging, as: (1) during RL training, there’s currently no source of truth; (2) training the model to be more cautious causes it to decline questions that it can answer correctly; and (3) supervised training misleads the model because the ideal answer depends on what the model knows, rather than what the human demonstrator knows.
ChatGPT is sensitive to tweaks to the input phrasing or attempting the same prompt multiple times. For example, given one phrasing of a question, the model can claim to not know the answer, but given a slight rephrase, can answer correctly.
The model is often excessively verbose and overuses certain phrases, such as restating that it’s a language model trained by OpenAI. These issues arise from biases in the training data (trainers prefer longer answers that look more comprehensive) and well-known over-optimization issues.12
Ideally, the model would ask clarifying questions when the user provided an ambiguous query. Instead, our current models usually guess what the user intended.
While we’ve made efforts to make the model refuse inappropriate requests, it will sometimes respond to harmful instructions or exhibit biased behavior. We’re using the Moderation API to warn or block certain types of unsafe content, but we expect it to have some false negatives and positives for now. We’re eager to collect user feedback to aid our ongoing work to improve this system.
AI for the benefit of Humanity…excluding Zimbabwe and 44 other countries
OpenAI is a non-profit that has brought us the most exciting publicly available AI tools to date. The first was DALL·E 2, an AI that can create an image based on your description, for example, a sunflower wearing shades. The image it produces looks undeniably true to life even though it was artificially generated. Then there is ChatGPT which is fast turning into everyone’s internet companion for both work and play.
OpenAI’s mission is to ensure that artificial general intelligence (AGI)—by which we mean highly autonomous systems that outperform humans at most economically valuable work—benefits all of humanity.
The tech they are making is meant to benefit all of humanity however it is not made available to all humanity. At the moment a total of 45 countries are not supported by OpenAI and the reasons are not stated. What they do say is this, “Geographic diversity and broadly distributed benefits are very important to us and we are working hard to increase the number of locations we can provide safe access to. If your location does not currently support API access, please check back for updates at a later date.”
There is no reason why these countries are not supported or when these same countries will be added to the supported list. However, this is a common theme with US-Based services. Countries that exist on a list of threats to the US are denied access to these services. Someone actually asked ChatGPT why some countries are not supported by OpenAI and this is the response it gave.
OpenAl is a research organization that is based in the United States and is subject to U.S. laws and regulations. There may be certain legal or regulatory restrictions that apply to OpenAl’s operations in certain countries, which could limit its ability to offer its services or products in those countries. In addition, there may be technical or logistical challenges that make it difficult for OpenAl to operate in certain countries. It is also possible that OpenAl has chosen not to offer its services in certain countries for business or strategic reasons. It is worth noting that the availability of OpenAl’s products and services may change over time, so it is always a good idea to check with OpenAl directly for the most up-to-date information.
When it comes to financial services we are used to this treatment in countries like Zimbabwe but with a service like ChatGPT which we just bundle up with services like Amazon Alexa and Google Assistant, it’s still a bit confusing as to why the service would be made unavailable in these countries listed below.
- Central African Republic
- Congo, Democratic Republic
- Equatorial Guinea
- French Southern Territories
- Heard Island & Mcdonald Islands
- Hong Kong
- Iran, Islamic Republic Of
- Lao People’s Democratic Republic
- Libyan Arab Jamahiriya
- North Korea
- Russian Federation
- Saudi Arabia
- South Sudan
- Syrian Arab Republic
- Viet Nam
How to use ChatGPT in unsupported regions
If you get the ‘Region is not supported’ notification then it’s going to be tricky to be able to use the app. A VPN can help with getting access to the signup page where you put in your email and password. Once you pass that stage you will need to put in a phone number for SMS verification. And if your phone number is from a country that is not supported by OpenAI you will not pass that stage.
So on top of a VPN, you can either ask a friend in a country that is supported by OpenAI to use their number so they get the SMS verification code or you can purchase a virtual phone number for a supported region through a service like Numero E-Sim.
It’s a bit of a faff for us in regions where the service is not available but if you really need to make use of this service then this might be the next best option as opposed to waiting indefinitely for whenever OpenAI eventually deems us worthy.
ChatGPT is being described by its creators as an artificial general intelligence meaning the scope of its application is wider than other implementations of AI we are used to interacting with on a daily basis. Assistants found in smartphones are generally limited to answering questions as briefly as possible and executing commands within the scope of the Assistant, the device it is running on, and the devices connected to it.
Smartphone CPUs are also now coming with sections of them dedicated to AI and the scope of this AI is limited to image and video post-processing (computational photography), and software optimization based on the device user’s habits to extend battery life without compromising user experience.
OpenAI’s ChatGPT is a bit different in that it is essentially all the internet but in chat format. The scope of it is pretty general and much less focused on specific objectives as is the case with other forms of AI mentioned above.
Socially, it’s a natural way of conversing and the ability it has to contextually carry on a conversation has made it a virtual buddy that one can talk to about virtually anything. Football, Formula 1, or fashion, it’s got knowledge of any social topic you can choose from. It may not always be factually accurate but that is how social interactions tend to be.
Academically it’s being used extensively with a lot of students as a way of more efficiently solving problems. Students are running their assignment questions into the AI and getting very decent answers. The same goes for essays and papers and the interesting part is that what ChatGPT spits out is not plagiarised content meaning it will pass plagiarism detection tools colleges and universities are currently using…until colleges and universities have tools to detect AI-generated scripts. Talk about letting the tech work for you.
The workspace is not spared as well. Developers left, right and center are relying on ChatGPT to quickly spot bugs in their code and massively cut down on the time they spend debugging and combing through forums for a solution. Content creation as well is having a bit of a shift. Ideas for articles or full articles can be produced by ChatGPT. The same goes for script ideas or full scripts for videos which the AI can also generate in a naturally sounding format that is very hard to tell from content authored by a human.
Google sees ChatGPT as its biggest threat right now
Google’s rise to fame is its search engine. When they came into play they provided the fastest and most comprehensive search engine which grew to the 92% market share they enjoy right now. They build a solid ad and e-commerce business on top of search if you are to search for something on Google right now, at least one of the top results will be an advert. These ads alone constitute 88% of Google’s revenue.
ChatGPT can pretty much be used as a search engine that brings relevant and simple results, much less of a hassle compared to Google’s list of links. So as a tool for search, ChatGPT is currently a more useful tool unless you need to reference the actual website. Google was already working on its own chat-based AI they were calling LaMDA. In fact, some of the learning models being used by ChatGPT are based on LaMDA.
Google had not released its AI because of the same issues that plagued Microsoft and Meta’s AI chatbots. They all use information from the internet when generating responses and some of this information might be factual whilst some may be fiction. As a result, Microsoft and Meta’s bot’s ended up releasing hate speech and racist comments which resulted in their demise. Google cannot let such an activity threaten its brand reputation that it has worked so hard to build.
The existence of ChatGPT presents a new threat to the search giant which is their biggest revenue earner. Search Ads. And it was enough of a threat for Google to issue an internal ‘Code Red’ on the matter with an internal memo citing that some staff was asked to abandon projects they were working on so that it’s all hands on deck to figure out Google’s alternative to OpenAI’s ChatGPT and roll it out as soon as possible.
Smaller companies like OpenAI can take some risks that bigger companies like Google, Microsoft and Meta can’t and get away with it. And that is why ChatGPT is available right now. It’s definitely going to change things and some of those changes might include how we use the internet. It’s not going to be a one-stop solution. It’s chat-based so any task that chat cannot handle this AI won’t be able to provide a solution for it. However if your biggest headaches can be solved in a chat then what a time to be alive!
- The LastPass hacking is VERY big. What’s the best way to manage passwords?
- The crazy tech at the Qatar 2022 FIFA World Cup
Quick NetOne, Econet, And Telecel Airtime Recharge
The post OpenAI claims the AI, ChatGPT, is meant to benefit all humanity but it is unavailable in 45 countries appeared first on Techzim.