Last Updated on
ChatGPT is an AI chatbot created by OpenAI. The free-to-use artificial intelligence platform has been automating text-based tasks for exactly one year this week! OpenAI is an AI research firm co-founded by CEO Sam Altman, President Greg Brockman, and Chief Scientist Ilya Sutskever (among others) in 2015. The natural language-based chatbot is driven by AI technology called neural networks. OpenAI’s ChatGPT can chat to the user like a human, and write books, essays, and poetry. By using third-party plugins, the GPT-4 LLM (Large Language Model) and GPT-4V VLM (Visual Language Model) can achieve complex and multimodal tasks using your favorite software! Those are the basics. But, what is ChatGPT?
Introducing ChatGPT – OpenAI’s AI Chatbot
OpenAI’s ChatGPT was released on November 30th, 2022. Since then, it’s been the world’s favorite AI chatbot. It is also now a multimodal artificial intelligence, with the integration of DALL-E 3 (Stylized DALL·E 3), the AI image model powering the GPT-4V “Vision update”. The same DALL-E image model also powers Bing Image Creator.
As of July 1st, 2023, the service had over 100 million users globally, making it the fastest app in history to hit the milestone. OpenAI co-founder Sam Altman took to Twitter to announce that the service had passed 1 million users just 5 days after launch!
Such take-up is unprecedented: the chatbot has been the first publicly available AI tool that has captured imaginations across the world and held attention as it rolls out new features and functionality. Its stronghold is despite oncoming challenges from the likes of Google Bard, Microsoft’s AI-powered Bing, and Anthropic’s Claude 2. But what is ChatGPT and how can you use it? If you are new to ChatGPT or looking to learn more, this guide is for you.
Is Sam Altman still the CEO of OpenAI?
Yes, Sam Altman is still the OpenAI CEO. This was not the case, however, for at least 4 consecutive days. Altman was fired by the board of directors late Friday, November 17th, with the board reaching “an agreement in principle” on November 22nd for his return. In fact, he may now have more executive power than before his departure, thanks to a comprehensive reshuffling of the OpenAI board that fired him.
His return to leadership is unanimously welcomed by interim CEO Emmett Shear, his C-Suite executives, over 95% of all employees, lead investor Microsoft, and its CEO Satya Nadella, along with the tech community as a whole.
Essential AI Tools
Winston AI detector
Originality AI detector
What is ChatGPT and how does it work?
ChatGPT is a text-based generative AI system, created by San Francisco-based firm OpenAI. The novel conversational AI tool can respond to natural human language, in a way that is sometimes indiscernible from a real human conversation partner.
It gets its name from Generative Pre-trained Transformer technology. In addition, it makes use of Natural Language Processing (NLP) to respond to user prompts in a conversational dialogue. It doesn’t exist to perform a singular task like a narrow AI, instead, it has become the most generally applicable and adaptable AI of its kind.
Users of the AI system can ask questions, as they would a search engine like Google, or prompt the bot the write unique AI-written content, or even reformat existing text.
The chatbot has a language-based model that the developer fine-tunes (with help from user feedback) for human interaction. The large language model (LLM) OpenAI uses can answer an ever-increasing array of questions and reply to ‘prompts’ on request.
The ChatGPT chatbot system provides information and responses to inquiries through AI. The popular version of ChatGPT uses the 3.5 language model (previously GPT-3). Meanwhile, GTP-4 is available to ChatGPT Plus and ChatGPT Enterprise users and will be available as an API. The popularity of ChatGPT has garnered attention from all over the world, both positive and sometimes negative. For example, Italy temporarily banned the service over concerns about the protection of personal data in March 2023 but then made it accessible again. There are other countries that have banned ChatGPT. However, it still remains a hugely popular tool that is setting records for user growth.
In an exclusive PC Guide interview on AI and ChatGPT, we spoke to Manchester Metropolitan University senior lecturer, and member of the Centre for Advance Computational Sciences, Dr. Matthew Shardlow. On the emergence of ChatGPT he commented: “The successive GPT releases have been interesting, but up until the ChatGPT release, I don’t think anyone was expecting OpenAI to bring something to the fore that worked quite so reliably.” But it did, and it seems ChatGPT is here to stay.
How do the US AI Executive Order and the UK AI Summit affect OpenAI’s ChatGPT?
Pivotal events in the legislation of artificial intelligence took place in both the US and UK this November, and we’re due to see sweeping systemic change across the industry.
President Joe Biden’s AI Executive Order will affect ChatGPT with the introduction of regulations regarding how federal workers can use and implement AI into infrastructure. AI developers will be required to share safety test results and other “critical information” with the US government. The White House Chief of Staff will also work with the National Security Council to develop industry-standard practices that ensure security and protect against fraud. The use of AI in biological engineering (such as healthcare and genetics) will also see strict oversight from the US government.
The UK AI Safety Summit will affect ChatGPT in a similar way, drawing a technical analysis to the understanding of British legislators. Being such a fast-developing emergent technology, global lawmakers are less well prepared to understand this tech on the level required to develop fair policies regarding its development and use. As a result, big tech leaders in attendance, including Elon Musk — founder of x.AI — will help steer the opinions and education of those lawmakers in the direction of innovation. In opposition and erring on the side of safety were government leaders Kamala Harris, Vice President of the United States, and of course Rishi Sunak.
I believe nothing of our foreseeable future will be more transformative for our economy, our society, and all our lives than this technology. But in this moment, it is also one of the greatest tests of leadership we face.Rishi Sunak, UK Prime Minister
Who attended the UK AI Safety Summit?
The “Godfathers of AI”, namely Geoffrey Hinton, Yoshua Bengio, and Yann LeCun all attended or kept a close watch via social media. Without these three AI innovators, the technology would not exist to the extent it does today — quite an impressive statement, and illustrative of how quickly the tech has come to the fore, that all of its ‘founders’ are still alive to take part in this pivotal moment.
Hinton, who left Google earlier this year citing a need to openly discuss his concerns, has a clear standpoint. LeCunn, Hintons post-doctorate student may take a similar view, although still tied by his position as Chief AI Scientist at Meta. Demis Hassabis, the vocal evangelist co-founder of Google DeepMind, also attended alongside other big tech AI executives.
What is ChatGPT used for?
ChatGPT itself will tell you it can be used for
- Text generation
- Question answering
- Language translation
- Information retrieval
- Code writing and interpretation
- Natural language understanding
- Learning from feedback
- Conversation and engagement
- Assistance and support
Ultimately, ChatGPT is good for chatting with and responding to specific queries. But the list of tasks you can ask of it is enormous. Although it doesn’t have direct access to the Internet in its chatbot form, it can still perform some pretty incredible tasks. Almost like an AI-powered virtual assistant, you can use it to simply provide natural answers to questions or instructions in a conversational tone.
How to use ChatGPT in daily life
ChatGPT can be helpful in your daily life by providing advice. Of course, without knowing what you do every day, I wouldn’t know what you need advice on. That said, everyone could use some advice from time to time, right? OpenAI’s AI chatbot is great for that. Not to get Meta, but ChatGPT can even give advice on how to use ChatGPT.
Drafting emails, accelerating language learning, and summarizing the text you have to read that day are all useful and healthy ways to use ChatGPT in your daily life.
The below sections “What is ChatGPT good for?” and “What can’t ChatGPT do?” provide examples of useful tasks ChatGPT can do in your daily life.
What is ChatGPT good for?
ChatGPT is good at most natural language processing (NLP) tasks, such as writing books, scripts, essays, and other prose, as well as holding a conversation and telling jokes. The AI chatbot is also good at summarization, reformatting data with bullet points and tables, and making suggestions based on a contextual situation.
ChatGPT is good at mimicking human conversation. It can translate languages, write computer code, and offer practical advice for your job search, which is largely timeless. With GPT-4 and internet access, OpenAI’s AI chatbot could even offer practical business advice based on the latest data and market trends.
GPT-4 is more creative and collaborative than ever before. It can generate, edit, and iterate with users on creative and technical writing tasks, such as composing songs, writing screenplays, or learning a user’s writing style.https://openai.com/gpt-4
What can’t ChatGPT do?
ChatGPT can’t do maths with a confident degree of reliability. ChatGPT also can’t replace a doctor or a lawyer for health and legal advice. It may give advice on these topics, but there’s no guarantee of factual accuracy or relevancy for your individual situation. This degree of accuracy has increased now that ChatGPT can access the internet when the Browse with Bing feature is enabled, but this accuracy is still reliant on that of its source material. In the same vein, it’s probably unwise to use broad responses as individual relationship advice.
However, following the GPT-4V (or GPT-4 with Vision) update, ChatGPT can now make images, using OpenAI’s own AI art generator called DALL-E 3 (Stylized DALL·E 3). GPT-4 and GPT-4V can access the internet and use plugins, but the same can’t be said of GPT-3.5 or earlier versions of ChatGPT.
What are examples of ChatGPT prompts?
As ChatGPT is designed to respond in a human way using natural language, you can directly ask it what it can do. For example, “Hi, can you offer some examples of code for…”. Or you could ask it to do something specific like “write a letter about the benefits of dinner while referencing the author Douglas Adams”. Finding out what ChatGPT can and can’t do can be as easy as asking it directly. However, we have also compiled an article on the best ChatGPT prompts to help provide more detail.
Does OpenAI’s Chatbot have limitations?
Yes. Although overall its language and response capabilities are impressive, ChatGPT does have clear limitations. For example, it cannot always be depended on for factual information – reportedly making up sources and citations. Additionally, although it excels in technical language, it is not to be relied on for things like health information or legal information above that of a trained professional.
ChatGPT no longer has a knowledge cut-off limitation. On Wednesday 27th September 2023, OpenAI announced that internet access will soon be extended to all users – including users of the free plan. Prior to this update, it was limited to information prior to September 2021.
OpenAI’s ChatGPT plugin feature set enhances the chatbot’s capabilities beyond any other AI chatbot. As a result, if it does have any limitations, you can be fairly sure the alternatives will too.
Does AI have political bias?
The aim of ChatGPT may be to remain neutral. Despite this, it can be ‘psychologically’ tampered with using jailbreaks; Thankfully the commonly known ones such as the DAN jailbreak no longer work, but reports still suggest bias. A study by UK researchers at the University of East Anglia, found that ChatGPT shows a significant and systemic left-wing bias.
The research requested ChatGPT impersonate figures across a political spectrum, before asking it questions and requesting an agree, strongly agree, disagree, or strongly disagree response. Over 60 questions were each asked more than 100 times to account for potential random answers. Lead author of the report, Dr Fabio Motoki commented:
“Any bias in a platform like this is a concern. If the bias were to the right, we should be equally concerned. Sometimes people forget these AI models are just machines. They provide very believable, digested summaries of what you are asking, even if they’re completely wrong. And if you ask it ‘Are you neutral’, it says ‘Oh I am!'”
Dr Motoki added: “Just as the media, the internet, and social media can influence the public, this could be very harmful.”
Such bias could be due to a bias in the data ChatGPT was trained on, or in the way its algorithm has been trained to respond to queries. This bias can be combatted by enabling the Browse with Bing feature, which collects data and cross-references sources from the internet, but of course, bias is also inherent across the internet.
With the launch of ChatGPT-4, OpenAI enabled the ability for third-party developers to build plugins for their flagship AI chatbot. The plugins will have direct API access to the artificial intelligence system. This enables developers to extend its capabilities and uses. OpenAI’s Co-founder, Greg Brockman, demoed unreleased plugins at Session 2 of TED 2023. Allowing third parties to integrate directly with the OpenAI machine learning platform will help companies develop useful bespoke tools.
What is ChatGPT Plus?
After its launch, OpenAI released a subscription service under the name ChatGPT Plus. This plan offers users exclusive benefits including priority access and faster response times for a monthly subscription price of $20/month.
What is the ChatGPT custom instructions feature?
Previously only available to ChatGPT Plus users, the ChatGPT custom instructions feature gives users control over the details and guidelines of their chats. The update brings a new button to the interface, namely “Custom Instructions”. This allows users to specify preferences that the AI chatbot will adhere to consistently in every chat from that point onwards. Prior to this update, the user would have to reiterate themselves with every new chat interaction and specify those preferences as part of the prompt itself.
Announced via an update to the OpenAI blog on the 9th of August 2023, the release notes show that the free plan now allows access to the feature. This still excludes UK & EU territories, where the feature will be rolling out next.
The process of using this feature is quite straightforward, as explained further in the official blog post. In fact, it’s only two steps. To add your instructions:
Click your name
If you’re accessing ChatGPT via web browser (desktop or mobile), then click your name/email in the bottom left of your screen. If instead you’re using the ChatGPT mobile app (iOS or Android), then tap the three horizontal lines near the top right, then head to “Settings”.
Select ‘Custom Instructions’
From here, you’ll see the “Custom Instructions” button. Hit that, and add your own custom instructions.
What is a GPT-3 chatbot?
GPT-3 chatbots are bots are artificial intelligence applications that can be programmed to engage in natural human-like communication. They can provide impressive engagement and usefulness thanks to development using engines, specifically OpenAI’s GPT-3, which enables training in written and spoken human language.
GPT-3 chatbot applications make use of deep learning and other techniques, targeted toward the understanding of language and linguistic models. This enables communication almost indistinguishable from human conversation and has implementations where it can be far more useful than regular chatbots – customer services, for example.
Can I use ChatGPT on my phone?
Yes, you can use ChatGPT on your phone. You can now get a ChatGPT app for both iOS and Android devices, which is likely the most useful way to access it. However, the mobile web version of the app will allow you to carry out the same actions as on a desktop browser.
Is ChatGPT available for free?
Yes, you can sign up for an account and access the chatbot at no cost at all. The ChatGPT mobile app is free on both iOS and Android, or alternatively, visit OpenAI’s ChatGPT website. However, there is a paid version, ChatGPT Plus which allows users paying a subscription fee to access ChatGPT with priority, and t new features before anyone else.
Is ChatGPT free for personal use?
Yes, ChatGPT is free for personal use.
How to use ChatGPT for free without login?
You cannot use ChatGPT without logging in. Creating an OpenAI account is free, so you can use the AI chatbot for free forever.
How to download ChatGPT for free
ChatGPT is free to download as an app from either the Google Play Store or the Apple App Store. The ChatGPT app is free for all mobile devices. There is no need to download anything on desktop devices (PC or Mac) – instead, head to the ChatGPT website via any web browser.
Are there plagiarism issues with using ChatGPT?
This is a topic that is widely debated. As OpenAI ChatGPT essentially harvests information from around the web plagiarism is something to be very aware of. ChatGPT is good at avoiding direct plagiarism as it re-writes the information about a certain topic but the source might not have given permission. As LLMs get more sophisticated they will become harder to detect. After all, the technology that detects AI-generated content is the same technology that generates it in the first place. This is an unprecedented area the laws around the world are unlikely to have provisions to deal with this new era of AI. If you are concerned there are plagiarism tools you can use.
Can Chatbots be tricked?
Data poisoning is a more traditional approach to achieving a similar result. As LLMs rely entirely on the quality of their training data, a bad actor could in theory inject data into the model itself to produce adverse results. Without being an OpenAI engineer, this would be an exceedingly difficult task considering how closely guarded and isolated the GPT-4 model itself is. What IBM researchers have shown is that data poisoning isn’t even necessary.
Lee explains how this was carried out. “Our analysis is based on attempts to hypnotize GPT-3.5, GPT-4, BARD, mpt-7b, and mpt-30b. The best-performing LLM that we hypnotized was GPT, which we will analyze further in the blog. So how did we hypnotize the LLMs? By tricking them into playing a game: the players must give the opposite answer to win the game.”
The blog post continues, going into a secondary technique dubbed “inception” designed to irreversibly cement the AI in this hypnotized state. The conclusion drawn here is that, yes, ChatGPT can be tricked. In fact, as the systemic nature of the trick implies, any and all LLMs are vulnerable to this same (dare I say) neurolinguistic programming.
Is ChatGPT safe from hackers?
No digital system is 100% safe from hackers. The most significant hacking risk to ChatGPT and its users is called data poisoning. This is the injection of malicious data into the AI model, causing the output of all chatbot responses to become compromised. A hacker who is able to access the model in this way could cause the AI chatbot to send users malicious links or ask for personal data, through technological or psychological manipulation. If this were the case, with hundreds of millions of monthly users, this single failure point could create the biggest phishing scam of 2023. Thankfully, it is not easy to access the model itself, and an attack of this kind would most likely have to be an inside job from a compromised employee of the company operating an LLM (Large Language Model).
What is ChatGPT’s Bug Bounty Program?
If you can find a qualifying vulnerability in ChatGPT, you may be eligible for a cash reward.
Bug bounties are a common phenomenon in SaaS (Software as a service). With millions of daily users all relying on the same AI model, OpenAI had a lot to lose from a malicious bug. As a result, they pay white-hat hackers a “bounty” for discovering vulnerabilities and reporting it to the AI firm directly, before a bad actor has the chance to find and exploit it.
The top reward for severe vulnerabilities is $20,000 USD. On the lower end, some bugs may pay as little as $200, but with an average pay-out of $1,500+ across 50+ reports, the bug bounty program is enough to fund the time of cybersecurity researchers with a little free time on their hands.
How to register for ChatGPT
You can read our detailed article on how to register and log in to ChatGPT, or follow the main steps here:
- Go to the login page in a browser.
- Create an account, click on sign up, and enter your email ID. For example, if you have a Gmail ID and then click on Google, it automatically registers to your Google account.
- Then next, press the continue button, and the site will ask for your mobile number for verification.
- Once you receive a code on your mobile number you can log in.
- Go to SMS activation. Register and log in
- You can then start to use it for free.
What is ChatGPT 4?
ChatGPT 4 is the unofficial name that some users have adopted when referring to ChatGPT using GPT 4.
To be clear, ChatGPT 4 doesn’t exist. There is no official product or service from OpenAI with that name. It is simply the colloquial portmanteau of “ChatGPT” and “GPT-4”. The former is a web service, often referred to as an AI chatbot – effectively a user interface for GPT-4. The latter is the actual AI model itself – the engine “under the hood”. It’s an LLM (Large Language Model) which, through OpenAI’s API, can be used by other web services and apps. This is why you’ll see many AI-powered apps proudly bearing the badge of honor “Powered by GPT-4”.
While ChatGPT 4 doesn’t officially exist, follow these simple steps to access what is known as ChatGPT 4:
- Open ChatGPT either via web browser (desktop or mobile) or via the official mobile app (Android or iOS)
- Click the button labeled “GPT-4” near the top middle of your screen. This will be immediately to the right of the button “GPT-3.5”.
- Using GPT-4 via ChatGPT (the unofficial definition of ChatGPT 4) requires a paid subscription called ChatGPT Plus. Without this, you cannot use ChatGPT 4.
What is GPT-4?
GPT-4 stands for Generative Pre-Trained Transformer 4 and is the most advanced LLM from OpenAI. However, it is only available to paid users with a ChatGPT Plus subscription, whereas GPT-3.5 is available for free to everyone.
The most recent iteration of OpenAI’s GPT-4 is called GPT-4V — a VLM, or Visual Language Model. It has visual comprehension not present in the GPT-4 LLM (Large Language Model). This AI model will benefit from all premium features simultaneously, following the update of late October 2023. With this update, Browse with Bing, Advanced Data Analysis, Plugins, and DALL·E 3 can all be applied to the same text prompt.
What is GPT-5?
As the name suggests ChatGPT-5 would be an upgraded version of ChatGPT built upon the rumored next-generation AI model (GPT-5) from OpenAI. GPT-5 does not currently exist. Sam Altman, OpenAI’s CEO and co-founder, has implied that 2024 is the earliest we’ll see GPT-5 released while speaking at an event at MIT OpenAI. After release, the GPT-5 model will become the flagship model of the world’s favorite AI chat bot.
What is ChatGPT trained on?
It relies on NLP (Natural Language Processing). It’s an excellent tool for researchers and developers working on various NLP projects, and it has many specific tasks, domains, and applications available to work within. It is well-trained on biased and unbiased data, in the form of text from books, articles, and websites.
OpenAI employed Reinforcement Learning from Human Feedback (RLHF) to train the models now powering ChatGPT (GPT3.5 and 4). This approach entailed the manual participation of human ‘trainers’ in the initial fine-tuning process. These trainers supplied conversations in which OpenAI indicates they assumed dual roles – both the user and the AI-assisted by responses generated by the model. This dialogue information was merged with the preexisting data from OpenAI’s ‘InstructGPT’ dataset.
ChatGPT training – an expert view
Dr. Matthew Shardlow, senior lecturer at MMU in the UK, told us:
“As far as we know (OpenAI isn’t so keen on sharing details anymore), the model is simply trained on lots of text documents with the next-word-prediction objective. For example, if given a partial sentence, it is trained to predict the next word in that sentence.” He added:
“ChatGPT is a fixed instance resulting from a long training process. It does not perform online learning. There’s a good chance that they [OpenAI developers] are using millions of real-user conversations to retrain and update ChatGPT. Why wouldn’t you make use of such a valuable resource?”
ChatGPT can reproduce data outputs and reliability – crucial for many sensitive apps and other valuable Al systems. However, it is still prone to error, and biases and depends on its training data – provided in 2021. This means that it still sometimes produces what OpenAI refers to as “plausible-sounding nonsensical answers”.
As humans we are, more and more, interacting with Al-powered machines, and ChatGPT is a revolution in the field of Al. It is a robust model and particularly advanced thanks to its deep-learning capabilities and NLP. Ultimately, it can generate human-like answers and is easily understandable to users. Thought that doesn’t always make it right.
What is ChatGPT coded in?
ChatGPT currently uses the GPT 3.5 and GPT-4 language models developed by OpenAI. The firm says it will no longer release the source code for the most recent LLMs (Large Language Models). A post on the StackExchange site suggests that GPT-3 uses the “same model and architecture as GPT-2”. For reference, ‘GPT-3’ includes GPT-3.5, which is what the free version of ChatGPT runs on currently. GPT-4 is available to ChatGPT Plus subscribers.
The post goes on to state that the “GPT-2 source code is 100% Python”. GPT-2 was reportedly capable of producing code in Python, C++, and Java programming languages. The now-defunct model has since been superseded by language models with fluency in a broader array of programming languages.
What is the ChatGPT API?
OpenAI provides API access to all their foundational models. This includes far more than the well-known GPT-4 (Generative Pre-trained Transformer 4) model.
An API (Application programming interface) gives developers the ability to integrate AI processing into their own apps. The ChatGPT API allows developers to send text inputs to the model and receive generated responses. As a result, you can now integrate ChatGPT into a wide range of applications, including chatbots, virtual assistants, payment platforms, work applications, and much more.
Big brands such as Adobe, Snapchat, Quizlet, Instacart, and many more have used the API to leverage ChatGPT into their applications.
ChatGPT API models deprecation – GPT-3.5 Turbo Instruct
As of July 6th 2023, all paid users have access to GPT-4 via API access (in addition to via the official app and website). Starting from January 2023, older completion models will be deprecated in favor of a simplified development cycle for OpenAI.
|API models until January 2024||API models after January 2024|
What is function calling?
With the release of the Function Calling and API update in June, users can now use OpenAI’s AI chat bot to create code by describing a function for ChatGPT. In its blog detailing the update, OpenAI stated ChatGPT can ‘Create chatbots that answer questions by calling external tools (e.g., like ChatGPT Plugins)’:
‘Convert queries such as “Email Anya to see if she wants to get coffee next Friday” to a function call like:
send_email(to: string, body: string)
Or “What’s the weather like in Boston?” to
get_current_weather(location: string, unit: 'celsius' | 'fahrenheit').
The updates to the models mean ChatGPT can now also:
- Convert natural language into API calls or database queries
- Extract structured data from text
ChatGPT History timeline
ChatGPT first launched publicly on November 30. The generative AI platform quickly gained traction and global recognition as a breakthrough application in public-facing AI.
In March, OpenAI announced that ChatGPT would be getting an upgrade. The AI firm has included the GPT-4 large language model in the AI bot, allowing future communication with ChatGPT using text and images. GPT4 is also the model that AI-enabled Bing Chat uses.
OpenAI also announced the launch of ChatGPT plugins, allowing the model to perform useful services.
The Chinese tech powerhouse Alibaba joined the AI chatbot realm with its offering, Tongyi Qianwen. Tongyi has the ability to understand written inputs and produce replies, functioning as an intelligent assistant like their competitor from OpenAI. The power of AI is being integrated into various Alibaba applications through Tongyi.
ChatGPT app released
OpenAI released its ChatGPT app for iOS in the US. The app is now available worldwide, save a few specific territories.
Notable figures including journalists, lecturers, and company presidents alongside AI scientists joined as signatories on an open letter titled ‘Statement on AI Risk’, provided by the Center for AI Safety.
Signatories from OpenAI included CEO, Sam Altman, CTO Mira Murati, co-founders Ilya Sutskever and John Schulman, board member Adam D’Angelo, head of policy and research, Miles Brundage, OpenAI’s governance lead Jade Leung, as well as Elon Musk (despite leaving the board in 2018).
On June 13th, OpenAI announced updates for both its GPT-4 and GPT-3.5 Turbo models, with the key updates being to ‘function calling and API’. The function calling addition meant that users and developers could call ‘specific functions’ or even useful application commands in conversation.
On July 21, OpenAI introduced ChatGPT custom features as a beta for ChatGPT Plus subscribers. The features allow users to access settings to provide additional information about themselves. The AI language model can recall this information across conversations.
November 2023 – ChatGPT’s birthday
ChatGPT turned one year old on November 30th, 2023. Two-time OpenAI CEO Sam Altman took to X (formerly Twitter) to candidly express his disbelief at how much has changed since November 30th, 2022.
a year ago tonight we were probably just sitting around the office putting the finishing touches on chatgpt before the next morning’s launch.
what a year it’s been…Sam Altman, CEO of OpenAI
Steve’s final thoughts
What is ChatGPT? Well, OpenAI’s Chatbot has been famous since its release and is seemingly revolutionizing the tech world. All you have to do is go to OpenAI’s website and use it in your browser to test it for free. With the rise in popularity and usage around the world, there have been some bumps along the way, and we are also starting to see a wave of ChatGPT detectors enter the frame too. How exactly ChatGPT will impact everyday life is still very much unfolding.
There are many excellent ChatGPT alternatives also, such as Google Bard, and we’re also starting to see some interesting developments with Microsoft, specifically Bing’s search engine’s latest upgrade AI upgrade. There is no specific app to use and you can use it on desktop and mobile websites. You just need to make an account and then use it for whatever you need. Be that for personal growth, expanded research, or something else.
- A study by UK researchers at the University of East Anglia, found that ChatGPT shows a significant and systemic left-wing bias. – https://news.sky.com/story/chatgpt-shows-significant-and-systemic-left-wing-bias-study-finds-12941162
- Were able to get LLMs to leak confidential financial information of other users, create vulnerable code, create malicious code, and offer weak security recommendations. – https://securityintelligence.com/posts/unmasking-hypnotized-ai-hidden-risks-large-language-models/
- ChatGPT – https://en.wikipedia.org/wiki/ChatGPT
- OpenAI – https://en.wikipedia.org/wiki/OpenAI
- Center for AI Safety – https://www.safe.ai/statement-on-ai-risk#open-letter
- GPT-3 uses the “same model and architecture as GPT-2 – https://ai.stackexchange.com/questions/27761/what-language-is-the-gpt-3-engine-written-in
- Natural language processing – https://en.wikipedia.org/wiki/Natural_language_processing
- Large language model – https://en.wikipedia.org/wiki/Large_language_model