A great AI tool announcement was made by browser provider Opera recently: “Opera Rolls Out Support For ChatGPT, ChatSonic, And AI Prompts.” I missed this announcement in my prior article on “Accessing AI: Answer Engines & LLM Chatbots” and the latest “This Week in AI” but I think it’s important enough to highlight it as its own article. Not just because this feature is worth trying out for yourself, but because it shows how search, AI and browsers are co-evolving into new features Answer Engines.
Opera announced the release of these AI features in its browser:
Shipping today in Opera and Opera GX are AI Prompts and sidebar integration of the popular GPT-based services ChatGPT and ChatSonic. Users can explore AIGC (AI-generated content) tools within Opera and Opera GX.
ChatSonic from WriteSonic is a bit like Bing Chat - “ChatSonic (now with GPT-4 capabilities), the conversational AI chatbot addresses the limitations of ChatGPT” including access to up-to-date information. The caveat is that, unlike free Bing Chat, getting access to the GPT-4-enabled version of ChatSonic requires a paid ($12/mo) account.
Opera also made clear this is just an initial step “with plans to move to stage two featuring its own browser AI engine in the near future.”
Using Opera for AI Features
Opera is a web browser and bills itself as fast, secure and user-friendly. To use these features, I had to download the Opera browser, just as I had to download Microsoft Edge to enable use of Bing Chat. Once downloaded, you need to set up Opera to enable the AI features. There are 3 AI-related features you get:
ChatGPT sidebar: This is a quick access to ChatGPT. It’s nice, but featurewise just a shortcut and you can bring it up in any browser on a new tab.
Chat-Sonic Sidebar: Because the GPT-4 version of ChatSonic requires paid subscription, I took a pass on this for now.
AI Prompts: These are prompt buttons you can use on any website you are visiting - “Shorten”, “Create a tweet”, “What’s the main point”. and more. I found this the most useful of the tools. This turns out to be an easier way to use chatGPT than a chat box for standard queries; instead of typing, I’m able to do standard queries with a single click.
This is for now a fairly thin layer of capabilities on the ChatGPT and other capabilities. Yet for people wanting to quickly synthesize data from a web page or document, the AI Prompts feature is a great time-saving utility. I do exactly this information extraction on web pages and documents constantly, and so this is enormously helpful. I’ve now found a use-case for AI to make browsing experience more useful, easier and better.
The Future of Search and Browsers in the AI Era
This offering from Opera tells us that Bing Chat is not a one-off. You can easily do similar things inside Chrome as well; for example, there are chatGPT Chrome extensions similar to Opera’s sidebars available. People will migrate to these tools for their utility.
Browser and search is at the front lines of how we will use Large Language Model-based AI, but it won’t end there for software. The rise of AI is a huge opportunity and challenge for all kinds of software:
AI means natural language can be the interface, displacing prior UI/UX forms
AI embedded intelligence can displace hard-coded optimization algorithms
AI content generation massively disrupts non-AI-based content editing software
AI transforms how people interact with information and data - search, writing, user documentation & manuals
AI disrupts practically every application across the entire software industry, and the Application providers have to figure out how to embed AI capabilities into each software application to keep that software relevant.
Opera made a good initial effort, but as they themselves note, it’s just a first step. Search engines will become more like Answer Engines and the browser will become a tool for getting answers not merely access to websites. Innovations will continue, and the only constant is change.
PostScript - Prompt Engineering 101
So my first AI prompt in Opera was on the web page “Prompt Engineering 101” and to prompt chatGPT “What’s the main point” to get a summary of it. The summary was a good one and is worth sharing as a result.
Prompt: What's the main point of "Prompt Engineering 101". Suggest keywords for further exploration.
Response:
Main Issue: The post explains the fundamentals of prompt engineering in LLMs and how it is the key to using them effectively.
LLMs: Large Language Models such as GPT-3, ChatGPT, AI21's Jurassic, and Cohere, have revolutionized what can be achieved with AI, and their pre-training on large datasets allows us to provide a textual prompt as instructions to complete a specific task.
Prompts: The instructions provided to an LLM are called prompts, and they are the input to an LLM, telling it what to do or how to think about a problem to get the most accurate output to a task possible. Prompt engineering is the process of adjusting a prompt to get specific and usable responses from an LLM, which is a key skill and the most significant part of using LLMs.
Best Practices: Prompt engineering can be a difficult task, but it is essential to get the most out of an LLM. The post covers best practices for creating prompts, which includes providing direct instructions, giving examples, and aligning prompt instructions with the task's end goal.
Keywords:
Large Language Models
Prompt Engineering
GPT-3
Few-Shot Learning
Best Practices