How AI Chatbot (ChatGPT) Is Shaping The Technology Industry
Artificial intelligence startup OpenAI launched ChatGPT, a chatbot that optimizes language models for dialogue to interact conversationally.
Bots are nothing new, but ChatGPT is unusually slick with language thanks to a training process that included digesting billions of words scraped from the web and other sources.
The dialogue format makes it possible for ChatGPT to answer followup questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests.
Its ability to generate short essays, literary parodies, and even functional computer code made it a social media sensation and the tech industry’s newest obsession.
Since the launch back in December, the future of education, the internet, and the tech industry have felt somewhat uncertain since.
What Is ChatGPT? How Does It Work?
ChatGPT took the internet by storm when it demonstrated an uncanny ability to answer questions and perform tricks like crafting mostly coherent essays, producing working computer code, and musing on the meaning of life.
It is powered by GPT-3, a text-generation algorithm developed by OpenAI, that has been fed huge quantities of text slurped from the web and other sources and then given additional training on how to answer questions.
This model using Reinforcement Learning from Human Feedback (RLHF), using the same methods as InstructGPT, but with slight differences in the data collection setup.
The initial model was trained by OpenAI using supervised fine-tuning:
- Human AI trainers provided conversations in which they played both sides—the user and an AI assistant.
- The trainers got access to model-written suggestions to help them compose their responses.
- The new dialogue dataset was then mixed with the InstructGPT dataset, which OpenAI has transformed into a dialogue format.
In order for OpenAI to create a reward model for reinforcement learning, they collected comparison data, which consisted of two or more model responses ranked by quality.
To collect this data, they took conversations that AI trainers had with the chatbot, randomly selected a model-written message, sampled several alternative completions, and had AI trainers rank them. Using these reward models, they were able to fine-tune the model using Proximal Policy Optimization.
ChatGPT is fine-tuned from a model in the GPT-3.5 series, which finished training in early 2022. ChatGPT and GPT 3.5 were trained on an Azure AI supercomputing infrastructure.
ChatGPT Limitations:
However, since ChatGPT works by finding statistical patterns in text rather than connecting words to meaning, it will often fabricate facts and figures, misunderstand questions, and exhibit biases found in its training data.
The sensitivity of the models to a simple rewording of the same question or misspelling in the prompts cause differences in responses,which is likely to complicate efforts to use this technology widely. It could end up mixing misleading and biased information into search results.
This makes the models unreliable for high-stakes use, such as translation in medical settings or content moderation, especially for those with marginalized identities.
The issues being spewed by the models today were seen as early as 2016, when Tay the chatbot was released, and again in 2019 with GTP-2.
- ChatGPT sometimes writes plausible-sounding but incorrect or nonsensical answers. To see the fix for this challenging issue click here.
- 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 and well-known over-optimization issues.
- Ideally, the model would ask clarifying questions when the user provided an ambiguous query but this model usually guesses what the user intended.
- While efforts have been made to make the model refuse inappropriate requests, it will sometimes respond to harmful instructions or exhibit biased behavior. Moderation API is being utilized to warn or block certain types of unsafe content, but expect it to have some false negatives and positives for now.
OpenAI stated on their website, “We know that many limitations remain as discussed above and we plan to make regular model updates to improve in such areas. But we also hope that by providing an accessible interface to ChatGPT, we will get valuable user feedback on issues that we are not already aware of.”
Users are encouraged to provide feedback on problematic model outputs through the UI, as well as on false positives/negatives from the external content filter which is also part of the interface.
They are particularly interested in feedback regarding harmful outputs that could occur in real-world, non-adversarial conditions, as well as feedback that helps them uncover and understand novel risks and possible mitigations.
The weight ultimately falls on them to not only obtain this feedback, but to handle the necessary tasks in the model building process before release, such as documenting, analyzing, and carefully curating data.
You can learn more about the ChatGPT 3.5 series here.
OpenAI’s technology is at the heart of interest in so-called generative AI, a term encompassing algorithms that can generate text, images, or other data.
The release of large language models like ChatGPT (a question-answering chatbot) and Galactica (a tool for scientific writing) has revived an old conversation about what these models are truly capable of.
Their capabilities have been presented as extraordinary, mind-blowing, autonomous; fascinated evangelists have claimed that these models contain “humanity’s scientific knowledge,” are approaching artificial general intelligence (AGI), and even resemble consciousness.
Some developers were so inspired by ChatGPT that they quickly used it to create apps, such as a spreadsheet assistant capable of performing complex calculations in response to a simple typed request.
One reason for tech industry excitement around ChatGPT, fed by its parlor tricks, is the suggestion it might disrupt the long-standing dominance of Google and other tech giants, by allowing small companies to outcompete much larger competitors. One popular theory is that the bot could transform web search.
Microsoft, is rumored to be weighing a $10 billion investment in OpenAI on top of an earlier $1 billion commitment.
When ChatGPT was asked how much the company behind it is worth the response included, “It is likely that its worth is in the hundreds of millions of dollars, if not more.”
Microsoft is betting that the company is worth a lot more, despite the fact neither ChatGPT nor other AI models made by OpenAI are yet raking in huge amounts of cash.
OpenAI has built several impressive and attention-grabbing demos and powers a popular autocomplete function for coders offered by Microsoft’s GitHub.
Microsoft also uses the technology behind ChatGPT to auto-generate snippets of code in its Visual Studio product, and it is reportedly looking at using ChatGPT to improve the performance of its laggardly search engine, Bing.
However, despite the hype swirling around its technology, the startup hasn’t created a breakout, highly lucrative product or business.
“We don’t really know what ChatGPT is going to be great at,” says James Cham, a partner at Bloomberg Beta, an investment firm, but Cham shares the feeling of many VCs and entrepreneurs that the technology behind the bot will pay out in a big way.
For Microsoft, investing more in OpenAI could offer a way to ensure that if there is a disruptive AI moment, the company will be not only unscathed but positioned to benefit.
Nathan Benaich, an investor with Air Street Capital who tracks trends in AI, says that while there is plenty of hype around generative AI, there are many ways that Microsoft could use the technology to improve its products.
The company’s broad and varied product range, from enterprise and cloud services to consumer devices and software, provide many potential opportunities.
ChatGPT is free for now, but investors are convinced the technology will become the basis of many new companies and products, perhaps including a new form of web search.
At the same time, it has a tendency to fabricate facts and repeat slurs or biases from its online training data. That has led to proposals to regulate the barely-launched technology.
The bot’s power has also prompted reflections on how humans are far from perfect and can spew bot-like BS of their own.
Furthermore, some schoolteachers have concluded that ChatGPT and other bots like it are here to stay; instead of trying to ban the technology, they’re embracing ChatGPT as a classroom tool to foster a more human-centric kind of learning.
Obraz na Sprzedaż
September 21, 2023Thank you for this helpful post! Your tips and insights have made a positive impact on my understanding of the topic. Keep up the great work!
Gotowe Ciasto
October 3, 2023Thank you for this helpful post! Your tips and insights have made a positive impact on my understanding of the topic. Keep up the great work!
Nieruchomości Alicante
October 8, 2023I wanted to express my appreciation for this post. It’s concise yet informative, and I’ve gained valuable insights from reading it. Thank you for sharing your expertise with us!
Maty Wygłuszające
October 12, 2023Kudos to you for creating such a great website! The content is top-notch, and your post is no exception. It provided me with the exact information I was looking for. Thank you!
caviar osciètre royal
October 16, 2023This post has been a great resource for me. It’s well-written and covers the topic comprehensively. Thank you for sharing your knowledge and expertise!
Pranie Fotela
October 20, 2023Thank you for this informative post! It has shed light on a topic I was struggling to understand. Your writing style is engaging and the information is presented clearly. Great job!
Women's jewelry
October 23, 2023Kudos to you for creating such a great website! The content is top-notch, and your post is no exception. It provided me with the exact information I was looking for. Thank you!
Edukacyjne Karty Pracy
October 28, 2023Your post is a goldmine of information! It’s evident that a lot of research and effort went into creating this valuable resource. Great job!
Webpanda
November 1, 2023Thank you for this outstanding post! It’s evident that you’ve put a lot of thought into it. The content is informative, engaging, and well-presented. Keep up the great work!
Travellernote
November 17, 2023Your post is incredibly helpful! It answered all the questions I had and provided additional insights. I’m grateful for the effort you’ve put into sharing this valuable information.
Kurs Nieruchomości
November 29, 2023This post has been a great resource for me. It’s well-written and covers the topic comprehensively. Thank you for sharing your knowledge and expertise!
Zakład Pogrzebowy Warszawa
December 6, 2023I wanted to express my appreciation for this post. It’s concise yet informative, and I’ve gained valuable insights from reading it. Thank you for sharing your expertise with us!
Healthcare
December 10, 2023I’m overjoyed that I discovered this website on Bing because it is exactly what I was looking for, and I’ve already saved it to my bookmarks.
Stacja ładowania EV
December 19, 2023Your post is a treasure trove of knowledge! It’s evident that you have a passion for the subject and have taken the time to compile valuable information. Thank you for your hard work!