ByteSize Daily Wrap - 2023/06/15

Hey there, Atometrix ByteSize readers, and welcome to yet another thrilling episode! We're beyond excited to explore the latest updates in AI, low code, and all things tech with you today. Join us as we delve into new improvements in AI language models, Google's stance on AI regulation, I-JEPA's future of efficient machine learning, and Amazon's secret AI plans!

Can you believe AI is improving so quickly that we can now create deepfake videos with just a few clicks? Programmers even have machine learning tools that learn like we do to spew out content faster than you can shake a selfie stick at. In a world where we rely on technology for everything from drinking coffee to traveling through space (okay, maybe not that last part... YET!), it's fascinating to see just how much these fascinating advancements are shaping our lives. Just try not to spill your coffee while marvelling at our techy future!

Going hands free? Now available as a podcast!

Have you had this forwarded to you? Subscribe now!

News & Articles

• (Link) OpenAI upgrades AI Language Models with Function Calling Feature: OpenAI has just launched its latest versions of AI language models, the GPT-3.5 turbo and GPT-4, and also launched a new feature called "function calling". "Function calling" enables developers to explain programming functions to the AI models, allowing them to execute those functions. The price for the GPT-3.5 turbo model has also been discounted by 25%. So go ahead, test their new features and save on costs.

• (Link) Google recommends multiple AI agencies for regulation: Google is advocating for multiple agencies to oversee AI tools, instead of creating a single regulator dedicated to the issue, unlike its competitors such as Microsoft and OpenAI. In a white paper, Google proposed that rules should be based on the intended use and potential risk of the AI, as a “one-size-fits-all” approach to AI regulation may be inappropriate. The company also recommended transparency in AI systems, including data used to train them. Google's opposition to a single regulator echoes concerns from the Trump administration about AI regulation impeding innovation. Did you hear about the robot who couldn't stop talking about its new app? It had a lot of interface.

• (Link) I-JEPA: The Future of Efficient Machine Learning: Meta has revealed a new computer vision model known as I-JEPA, which can learn more efficiently than previous systems. Similarly to human learning, the model can construct internal models of how the world functions, plan the best way to carry out complex tasks, and adapt to unfamiliar situations. It is anticipated that I-JEPA could help the machine learning sector develop at a faster pace, delivering better results in less time. "Looks like computers are getting smarter than us!"

• (Link) Amazon's Secret AI Plans Exposed: ChatGPT Takes the Lead: An internal Amazon document has been leaked that urges employees to use the AI language model ChatGPT for various tasks such as customer service, personalisation and content creation. The document emphasises the model’s accuracy and speed, and encourages teams to explore its potential uses. This reinforces Amazon’s increasing efforts to implement AI technology across various sectors of the company. Oops, looks like someone spilled the beans on Amazon’s latest AI plans!

Insights & Papers

• (Link) AI Improvement Technique unleashed in GAP: A new technique called Gradient Ascent Post-training (GAP) can significantly improve pre-trained language models by updating them with untagged random text in only a few steps. This allows smaller models to perform as well as larger ones without any task-specific adjustments. In other words, a small model using GAP can be just as powerful as a larger one. With such significant advancements in language models, soon AI will rise to a level where they will be able to tell a better dad joke than well... dads.

• (Link) 92% of Programmers Rely on AI to Boost Productivity: According to a recent survey, a whopping 92% of programmers use artificial intelligence (AI) to aid them in their work. From predictive analytics tools to natural language processing, AI technologies are becoming increasingly popular among developers. The survey also found that developers are most interested in learning more about machine learning and deep learning. Looks like the robots might be taking over after all! Just kidding, but it's clear that AI is quickly becoming an essential tool in the programming world.

• (Link) Financial Programming with Open-Source Models: The FinGPT project has released a paper and open-source financial language models (LLMs), allowing researchers to train on financial texts and assist with financial programming tasks. The models are trained to recognise financial words and phrases and can be adapted to recognise domain-specific language. With vast potential to accelerate financial data analysis and investment research, this project is set to make a big splash in the world of finance! Who needs human financial experts now?... Just kidding!

• (Link) Video Editing with Text-Driven Rerendering Tech!: The article is about a new technology called zero-shot text-guided video-to-video translation that allows for the rerendering of video content. Essentially, this means that with just text as a guide, video footage can be transformed into different content. The implications of this tech are pretty cool and could revolutionise video editing. Maybe next time you're filming, you could turn it into a sci-fi movie with the touch of a button!

Code & Engineering

• (Link) LLaMA-Tokenizer-JS: The Unbeatable JavaScript Tokeniser!: LLaMA-Tokenizer-JS is a JavaScript tokeniser that enables LLaMA to work seamlessly on websites and Node.js. Its code is available on the GitHub Repo, making it easy to install and integrate into existing projects. This tokeniser will help developers improve their web development process and provide a more seamless experience for their users. You'll be able to tokenise your LLaMA with ease using this tool! "Just llama-zing!"

• (Link) Transform Your Detection with Detrex: The Ultimate Toolbox!: Detrex is a new open-source toolbox that utilises leading-edge transformer-based detection algorithms. Boasting a range of superior features and capabilities, the program is built on top of Detectron2 and uses elements from both MMDetection and DETR. You could say it's the Transformer of detection algorithms!

• (Link) AlgoPerf: Deep Learning Algorithm Training!: Researchers have created AlgoPerf, a competitive benchmark system, to evaluate deep learning training algorithms and set a foundation for future enhancements. The system aims to improve data processing speed, save resources, and create accurate models. With AlgoPerf, researchers can compare and improve various training algorithms fairly. "Let's see which algorithm comes out on top!"

• (Link) Create Convincing Deepfakes in Seconds with Roop: Roop is a GitHub repository that enables the creation of deepfakes with just one click. Put simply, the tool makes it easy to edit videos and images with AI-generated content. With Roop, users can create fake images by swapping out faces, backgrounds and other details, essentially creating a convincing illusion. Be wary though, the technology can also be used maliciously. Just because you can create deepfakes quickly doesn't mean you should!

Tools & Products

• (Link) Supercharge Your Marketing with Assembly's Calendar: Assembly has just launched a new marketing calendar that meets the needs of marketers and exceeds their expectations. Unlike other content calendars, Assembly comes with features such as manager approvals, AI content generation and analytics integration. Marketers no longer need to hack together their own solutions using Notion or Sheets. It's like having your own marketing superpowers at your fingertips! Who needs a cape and a secret identity when you have Assembly?

• (Link) Revolutionary AI Tool Personalises All App Experiences: The article discusses Augment AI, a new AI tool that can seamlessly work across all apps, assisting the user in tasks such as scheduling, ordering food, and booking appointments. Unlike other AI chat services, Augment AI can learn the user's habits and make personalised suggestions and recommendations. The tool aims to reduce app fatigue and streamline user experiences. Imagine having a personal assistant across all your apps! This might be the end of having to toggle between different apps constantly.

• (Link) Generate Unique Articles in Seconds with Sqriblr & ChatGPT: Sqriblr is a tool that allows you to generate articles quickly and easily with the help of ChatGPT. You can generate unique articles with images in just seconds, making the process more efficient and time-saving. Say goodbye to spending hours researching and writing articles, and say hello to Sqriblr! With this tool, you can have fun while also getting quality content that will engage your audience. So what are you waiting for? Give it a try and see for yourself! And before you go, remember that a life without laughter is like a day without sunshine.

Videos

• (Link) Can AI achieve consciousness? Experts weigh in: The article discusses the topic of artificial intelligence (AI) and whether it can possess conscious awareness like humans. While AI has advanced significantly in recent years with machine learning and cognitive computing, experts argue that the current technology does not demonstrate true consciousness. Some believe that we may need to develop a completely new form of machine learning to achieve it. Until then, it seems that we still have a long way to go before AI can truly replicate human consciousness. Looks like we’re safe from our robot overlords for now.

• (Link) Create Your Own Website with AI: No Coding Needed!: The Framer Tutorial teaches you how to generate a live website using AI. With easy to follow steps, even those with limited coding experience can create a functional website. Say goodbye to the days of needing a team of developers to create your website. With this tutorial, you'll have a user-friendly website in no time. Who needs developers anyway? Just kidding, we all do… sometimes.

Miscellaneous

• (Link) Revamp your Web Scraping with ChatGPT Automation: The article provides a tutorial explaining how to use ChatGPT to automate web scraping. It outlines the steps involved in creating a custom chatbot, setting up the environment, and training the bot to extract information from web pages. Upon completion of the tutorial, the reader will be able to fully automate their web scraping tasks. So, say goodbye to manual labour and hello to ChatGPT automation!

• (Link) AI Lies Exposed: Algorithm Detects Fake Texts 99% of the Time: Researchers at the University of Kansas have developed an algorithm that can detect fake text generated by the AI language model GPT-3 with an accuracy rate of 99%. The team trained the algorithm on a sample of genuine and fake text from the model, which led to the detection of patterns in the generated text. The researchers hope their findings will help prevent the spread of fake news or misinformation generated by AI language models. Now the robots can no longer fool us with their fake texts, we're safe!

• (Link) AI discovers 3 new Nasca lines in Peru: Exciting news for archaeology enthusiasts, as AI has identified three previously unknown Nasca Lines in the Peruvian desert. The lines were discovered using machine learning algorithms that analysed satellite imagery of the area. Experts believe that the lines, which include a humanoid figure and two animals, were created by the ancient Nasca people between 200 BCE and 500 CE. This discovery highlights the potential for technology to aid in the discovery of new historical sites and artifacts. Who needs Indiana Jones when you have AI?

• (Link) Ikea Revolutionises Furniture Sales with AI and Remote Design: Furniture retailer Ikea is turning to remote interior design and AI-powered innovation to bring a new era for selling furniture. By taking the sales experience online, customers get the opportunity to consult with designers through virtual calls to create a home environment that matches their tastes. The move to the online sphere comes as the pandemic has accelerated the trend for online sales. To improve the sale strategy, Ikea is looking at AI to create products that meet customers' needs. Now, you no longer have to leave your home to get your favourite decor.

• (Link) AI helps Paul McCartney create new Beatles hit: Legendary musician Paul McCartney has been experimenting with AI technology to create a new Beatles song. McCartney used the tech to isolate John Lennon’s vocals and layer them with a never-released demo track to produce a fresh new composition that merges the distinct voices of the Beatles legends. Perhaps McCartney's next project will be an AI-powered Beatles reunion tour!

Art of the Day

_{“Mermaid sitting on a coral reef, white background, watercolor clipart.” By @duchuynh2308}

Well, ByteSizers, we've reached the end of today's juicy tech adventure! Thanks for joining us for another daily dose of AI, low code, and tech updates. Remember, even when the sun goes down, keep shining bright and stay ByteSized! Catch you on the flip side!

p.s. Don’t forget to checkout our platform or our apps!

• Atometrix: The only low code platform the allows easy deployment of server side React Native. Now you can edit your UI and deploy changes without a developer and see it happen in your production apps!

• Storyworm: The world’s first server side React Native application. Storyworm is a dynamic platform for all storytellers. Powered by AI, it lets you craft unique narratives, voice them through a text-to-speech feature, and combat writer's block with AI-generated plot suggestions. Beyond being an app, Storyworm is an inclusive community for story-lovers worldwide. This tool enables an immersive narrative experience that adapts and evolves according to users' needs.