Dev News: GPT-4 Turbo, Chrome Talks Pretty, and Worlds Merge
OpenAI announced several goodies for developers on Monday, including a new way to create personalized chatbots and lower costs for GPT-4 Turbo input and output tokens.
GPT-4 Turbo is the company’s latest AI model, with context up to April 2023 and more input capacity. Previously, GPT-4’s knowledge was cut off at Jan. 2022. The old model also limited users to 3,000 words, but GPT-4 Turbo can accept up to 300 pages in length. It also supports DALL-E 3 AI generated image and text-to-speech.
Currently, Turbo is available in preview for developers, but will be fully released in a few weeks. OpenAI said its GPT-4 Turbo input tokens are three times cheaper than GPT-4 at $0.01. Output tokens are two times cheaper at $0.03, CNBC.com reported Monday.
The big push for Generative AI right now seems to be creating personalized bots for businesses or users. No coding is required to build a customized GPT — it’s as easy as starting a conversation, giving the AI instructions and extra knowledge, and picking what it can do — with options like searching the web, making images, or analyzing data.
OpenAI also introduced its own App Store for people who create their own GPTs and want to make them available for public downloads. In the coming months, chatbot creators will be able to earn money based on their creation’s usage, CNBC reported.
In other news, ChatGPT became the target of attacks this week. OpenAI first reported seeing problems with its LLM-based chatbot and API on Tuesday, with a major outage reported Wednesday, when the company revealed it was the target of “an abnormal traffic pattern reflective of a DDoS attack,” Security Week reported.
Also this week, GitHub also added Duo Chat to its own AI lineup. Duo, revealed earlier this year, is a set of AI features to help developers be more productive. It summarizes issues and generates descriptions of epics and issues, as well as through code suggestions and vulnerability explanations. Duo Chat, which is in beta, allows developers to have a ChatGPT-like interaction with Duo.
Chrome Talk Pretty One Day
Chrome 117 supports a new wrapping feature that solves orphans. Widows are words alone at the top of a text block and orphans are alone at the end of a text block. They’re considered gauche in design and layout.
The new tag is:
CSS text-wrap: pretty.
It will also adjust hyphenation if consecutive hyphenated lines appear at the end of a paragraph, or adjust previous lines to make room. Finally, it will adjust for text justification.
Pretty is different from the CSS orphans property, which is relevant when using CSS multi-column layout, according to the Chrome team’s blog post.
Web and Desktop Worlds Are Colliding!
There are web apps and there are desktop apps, and never the twain shall meet. Well, that’s changing with new frameworks that can share code across both spaces, according to Sam Basu, a Microsoft MVP and Progress Developer Advocate for Telerik products.
In a recent DevOps Magazine article, Basu explored the ways in which .Net and JS developers can benefit from the emerging tools and platforms that support both.
“The desire to have a crossover is nothing new,” Basu wrote. “What is new is that frameworks, platforms and tooling have matured for developers, creating new possibilities to share code between web and desktop apps, with benefits for both .NET and JS developers.”
Observability Platform Adds Chatbot
LogicMonitor, which is a SaaS-based hybrid observability platform, launched LM Co-Pilot (not to be confused with GitHub’s Copilot). It’s a generative AI designed to assist users in “day-to-day operations, recognize issues and offer solutions” related to IT and cloud operations.
The generative AI system can take information and distill it into a refined, interactive experience, said Taggart Matthiesen, chief product officer at LogicMonitor.
“With Co-Pilot, we can condense multiple steps into an interactive experience, helping our users immediately access our entire support catalog at the tip of their fingers,” he said. “Co-Pilot minimizes error-prone activities, saves our users time, and exposes them to contextually relevant information.”