Opera jumped on Web3 quickly with its Opera Crypto Browser, and it seems like it will do the same with generative AI. Opera wants to add AI features to the sidebar of the browser. Opera is finally getting into the AI race by adding ChatGPT to its browser. The first thing to be tested is a new “Shorten” feature that will use an AI chatbot to summarise web pages and articles.
By adding more features, the company wants to improve the browsing experience. Microsoft has already started to add ChatGPT to the Edge browser and to the Bing search engine. On the other hand, Google has started building an AI-based chatbot called Bard. Opera is adding ChatGPT to the sidebar.
Table of Contents
What is ChatGPT?
ChatGPT or Generative Pre-trained OpenAI made Transformer, a chatbot made by OpenAI that had already been trained, came out in November 2022. It is built on top of OpenAI’s GPT-3 family of large language models and has been fine-tuned using both supervised learning and reinforcement learning, which is a form of transfer learning.
On November 30, 2022, ChatGPT came out as a prototype. It got a lot of attention quickly because it gave clear answers and explanations in many different areas of knowledge. On the other hand, its uneven accuracy with facts was seen as a big problem. After ChatGPT came out, it was thought that OpenAI was worth $29 billion.
Also Read: How To Use Your IPhone’s Notes App To Secretly Send Messages
Training
On top of GPT-3.5, supervised learning and reinforcement learning were used to improve ChatGPT, which is a generative pre-trained transformer (GPT). This is one type of transfer learning. In both cases, the model’s performance was improved by human trainers.
In supervised learning, trainers played both the user and the AI assistant in conversations with the model. During the reinforcement step, people first ranked the model’s answers from a previous conversation.
Using these rankings, “reward models” were made, which were then used to fine-tune the model several times using Proximal Policy Optimization (PPO). Proximal Policy Optimization algorithms are a cost-efficient way to improve trust region policy optimization algorithms.
They get rid of a lot of the things that take a computer a long time to do and make the ones that are left faster. With help from Microsoft, the models were trained on its Azure supercomputing infrastructure.
OpenAI also keeps getting information from ChatGPT users that could be used to train ChatGPT and make it better. The answers users get from ChatGPT can be voted up or down. They can also add more feedback in a text field when they upvote or downvote.
Opera is Integrating ChatGPT With Website and Article Summaries
All of Opera’s money is going into AI. This week, the company said that it wants to add AI that can make things on its own to its web browser. The first step will be “Shorten,” a feature that will use ChatGPT to summarise articles and webpages.
Once everyone can use the tool, a new icon will show up to the right of the address bar. If you tap it, ChatGPT will show you a sidebar with a bulleted summary of the page you’re looking at. The vice president of marketing and communications at Opera, Jan Standel, said that users will be able to use Shorten “very soon.”
The company is working on more AI-powered features that it says will “enhance” the Opera experience, but it hasn’t said what these features will be. Shorten was announced the same week that Microsoft said it would change Edge to make it have a “AI-powered copilot.”
The new Prometheus model from the company can, among other things, sum up web pages. Google also said last week that it is making Bard, an AI chatbot that will be run by its LaMDA platform. The timing of the announcements suggests that Opera and Microsoft want to take market share away from Google by using generative AI. But it’s still not clear if these changes will actually make people stop using Chrome.
Also Read: How To Fix Bridge.Sys Blue Screen In Windows? How Can We Fix It?
How to Use ChatGPT in the Opera Browser?
The company explained how the new ChatGPT feature works in a blog post. Opera users can turn on the new ChatGPT feature by clicking the “shorten” button to the right of the address bar. If you’re using ChatGPT, a summary of the page or article you’re looking at will show up in a sidebar on the left. Users will also be able to go straight to ChatGPT by using the sidebar.
Unfortunately, not everyone has access to the “shorten” feature yet. Jan Standel, Opera’s VP of marketing and communications, says that ChatGPT will “launch in browsers very soon.” AI-based features that make browsing better are also being made by the company.
Along with the “Shorten” feature for ChatGPT, Opera plans to add “popular AI-generated content services to the sidebar.” Users who get the Opera Browser Beta or developer build can get these AI ChatGPT features before anyone else.
Aspects of ChatGPT
ChatGPT can do a lot of different things, even though its main job is to talk like a person. For example, it can write and fix computer programmes, write music, teleplays, fairy tales, and student essays, answer test questions, write poetry and song lyrics, emulate a Linux system, simulate an entire chat room, play games like tic-tac-toe, and simulate an ATM.
ChatGPT’s training data includes man pages and information about Internet phenomena and programming languages, such as bulletin board systems and the Python programming language. ChatGPT is different from most chatbots because it remembers what it was asked before in the same conversation.
Journalists have said that this makes ChatGPT useful as a personalised therapist. The questions asked on ChatGPT go through OpenAI’s company-wide moderation API, which removes any questions that could be racist or sexist. This keeps ChatGPT from showing or making offensive results.
Also Read: How to Fix Waves MaxxAudio Service Application High CPU in Windows?
Problems with ChatGPT
There are a lot of things wrong with ChatGPT. OpenAI said that ChatGPT “sometimes writes answers that sound right but are wrong or don’t make sense.” This is a common trait of large language models, and it’s called “artificial intelligence hallucination.”
Goodhart’s law says that because ChatGPT is based on human oversight, if the reward model is too good, it can hurt performance. ChatGPT doesn’t have a lot of information about what happened after 2021.
The BBC says that ChatGPT users won’t be able to “express political opinions or engage in political activism” after December 2022. But research shows that ChatGPT leans left-liberal and pro-environment when two well-known voting advice apps ask it to take a stand on political statements.
Human reviewers preferred longer answers when training ChatGPT, no matter how well they were understood or how much factual information they had. There is also algorithmic bias in training data, which can be seen when ChatGPT answers questions with words that describe people. In one case, ChatGPT made a rap that said women and people of colour who were scientists were not as good as white men who were scientists.
