The Past Year in the Field of AI – An Annual Breakthrough

January 2023

– Anthropic Unveils LLM Generation 1: Claude: Anthropic introduces a new artificial intelligence chatbot named Claude, similar to ChatGPT. Based on research on the Constitution of Artificial Intelligence, Claude differs from ChatGPT in its approach to reinforcement learning and employs a preliminary model-building method to initially rank fine-tuned outputs instead of relying solely on human feedback. Claude emerges as a serious contender to ChatGPT, showcasing improvements in areas such as its ability to write coherently about itself and its limitations, and producing more natural-sounding text.

 – Microsoft + OpenAI = Azure OpenAI Service: : Microsoft announces an overview of the Azure OpenAI Service, a tool that enables businesses to access state-of-the-art artificial intelligence models, including GPT-3.5, Codex, and DALL·E 2. The service allows you to develop advanced applications, enhance user experiences, and optimize internal performance. It also includes ChatGPT running inference on the Azure artificial intelligence infrastructure.

 – VALL-E: Text-Based Language Model for Speech Synthesis: Microsoft releases VALL-E, a text-based language model for Text-to-Speech (TTS) synthesis. Trained on 60,000 hours of English speech, it can generate high-quality personalized speech using only a short recording of a reference speaker. It can also retain the emotions and ambiance of the reference recording in the generated speech.

February 2023

 – Google Launches Bard: Google introduces Bard, an AI-powered conversational platform fueled by LaMDA (Language Model for Dialogue Applications), to provide fresh and high-quality responses by pulling information from the web. The hype surrounding Bard is substantial, but it later diminishes due to high levels of imagination in its responses.

 – ChatGPT Hits 100 Million Users: OpenAI’s popular chatbot, ChatGPT, reaches the milestone of 100 million users just two months after its launch, with unprecedented consumer application growth rates, according to analysts at UBS. The app garnered 590 million visits in January from 100 million unique users, compared to TikTok, which took 9 months, and Instagram, which took 2 years, to reach the same milestone.

 – Meta Launches LLaMA : After Google and Microsoft, Meta joins the race in the field of Large Language Models (LLM) with its new release, LLaMA, a language model ranging from 7 billion to 65 billion parameters. LLaMA-13B surpasses GPT-3 (175B) on most metrics, and LLaMA-65B competes with top models like Chinchilla70B and PaLM-540B.

 – No Lights, No Camera, All Action: Runway introduces Gen-1, an artificial intelligence system that generates new videos from existing ones using images or text prompts. It relies on a structure and content-guided video synthesis model trained on images and videos, providing nuanced control over output characteristics. The platform offers various modes for styling, scripting, face swapping, and customization.

March 2023

 – OpenAI Launches GPT-4: OpenAI releases GPT-4, their next-generation large-scale multimodal model that can take input in the form of images and text, producing text output. Surpassing other large language models, GPT-4 excels in various language and reasoning tasks, demonstrating professional and academic-level performance.

 – Adobe Introduces Firefly: Adobe unveils Firefly, a creative artificial intelligence tool that enables users to generate images, add or remove objects, and transform text. Firefly includes various modules, such as Generative Fill, Generative Recolor, and 3D to Image. Integrated into Adobe applications like Photoshop, Illustrator, and Adobe Express, Firefly introduces new features supported by creative artificial intelligence.

 – Runway Quickly Transitions from Gen-1 to Gen-2: Runway introduces Gen-2, a multimodal artificial intelligence system capable of generating new videos using text, images, or video clips. It can synthesize videos in any style using text prompts, convert image styles to videos, and customize the model for higher quality results.

April 2023

 – Elon Musk Unveils xAI: Elon Musk introduces xAI for artificial intelligence research, positioning it as a formidable competitor to AI leaders like OpenAI, Anthropic, and Microsoft. He also acquires Twitter and rebrands it as X. Whether it truly transforms Twitter into a sophisticated and convenient platform remains a debatable topic!

 – Amazon’s Footsteps in the LLM World: Amazon unveils Amazon Bedrock, enabling the creation of innovative artificial intelligence applications through third-party models from AI21 Labs, Anthropic, and Stability AI, as well as AWS’s Titan FM models. Amazon also introduces CodeWhisperer, an AI-powered programming companion for swift and secure application development, available for free to individual developers.

 – Meta AI “Cuts” Through Any Object with AI: Meta AI releases the Segment Any Object Model (SAM), an artificial intelligence model capable of segmenting any object in an image with just a click, using input suggestions trained on over 11 million images with more than 1 billion masks.

May 2023

 – Stable Diffusion XL: Stability AI introduces Stable Diffusion XL (SDXL), the latest image generation model with the ability to reproduce images with exceptionally high fidelity, improvements in image structure, and facial generation. SDXL also includes text-based image generation requests that users can command.

 – Runway Gen-2.5: Runway updates its system to Gen-2.5, introducing several new enhancements. One of the key features is the ability to seamlessly transition between various video generation styles and support multiple data sources, including text, images, and videos.

 – Microsoft’s Upgrade for VALL-E: Microsoft releases an update for VALL-E with improved high-quality voice synthesis and customization options to match the user’s style. This update also introduces a unique feature allowing users to “listen” to the generated voice and make adjustments accordingly.

June 2023

 – Google Launches DALL·E 2: Google introduces DALL·E 2, the next version of its popular image generation model. DALL·E 2 boasts 2 billion parameters and has the ability to create high-resolution images with realism and significant diversity. It also includes the capability to generate animated images.

 – Adobe Firefly 2.0: Adobe announces version 2.0 of Firefly, bringing many new features and enhanced customization capabilities. The new version can generate images and videos with outstanding realism, offering unlimited creativity for users.

 – OpenAI Unveils GPT-4.5: OpenAI continues its development by launching GPT-4.5, an update between the GPT-4 and GPT-5 stages. This update provides enhanced language and multimodal performance, with the ability to understand and respond to users in a natural and flexible manner.

July 2023

 – Humane Releases AI Pin: A startup founded by former Apple designers, Humane, unveils the “AI Pin.” It’s a smart wearable device with an integrated projector that could potentially replace mobile phones in the future!

 – Meta AI Releases Llama-2 Open Source: Llama-2, a series of large language models (LLMs) with 7 billion to 70 billion parameters, is released by Meta AI, offering three sizes (7 billion, 13 billion, 70 billion). Trained on 2 million GPUs, over 40% more than its predecessor, Llama-1, and including over 1 million human annotations. With a 4k context length, Llama-2 outperforms other open-source models on diverse benchmarks, including programming and reasoning.

ChatGPT Receives Code Interpreter OpenAI has rolled out the Code Interpreter for all ChatGPT Plus users. It can execute code, analyze data, create charts, edit files, perform mathematical operations, and much more!

August 2023

Meta AI Releases Code Llama Meta AI introduces Code Llama, a sophisticated open-source large language model specifically designed for programming tasks. Built on the robust architecture of Llama-2, it has three variants to cater to various programming needs, including a specialized model for Python.

Clone Any Voice with AI in < 3 Seconds VALL-E X is a versatile multilingual model for Text-to-Speech (TTS) proposed by Microsoft. While Microsoft initially published a research paper, they didn’t release the source code or pre-trained model. However, we now have an open implementation of VALL-E X that you can use to clone any voice on your computer.

Google’s New Programming Platform: Google launches Project IDX, a browser-based cloud programming environment, showcasing advanced AI features like code translation and automatic code completion. This is not just about fast programming but also intelligent programming. While competitors like GitHub’s Copilot have ventured into AI-assisted programming, Google’s comprehensive approach is unique.

September 2023

Microsoft’s AI-Powered Smart Backpack: Microsoft has just introduced a smart backpack equipped with artificial intelligence, going beyond mere storage space. With the ability to see, hear, and interact with the environment around you, this backpack takes digital assistance to a whole new level.

Self-Driving into the Future Wayve introduces LINGO-1, an open-loop driving perception, leveraging natural language in autonomous driving to enhance explanation capabilities and model training. They draw inspiration from LLMs and combine language with image and action data to create VLAMs.

Full 360° Synthesis of 3D Heads Meet PanoHead, the game-changer in the world of creating 3D heads. Unlike other 3D models, PanoHead provides an impressive, consistent perspective from all angles, not just the front. The secret sauce? A fully detailed and remarkably clear 3D chart from every angle.

October 2023

XAI Makes Its Debut: In a significant move, leading companies like OpenAI, Google, and Meta AI begin integrating powerful explainability features into their AI models. This makes AI more transparent, helping users understand how the model makes decisions and enhancing trust.

Natural Language Controls for Drones SkyTalk, a new system from DJI, allows you to control drones using natural language instead of using a mobile phone or controller. It employs LLM models to understand and execute natural language commands, expanding how users interact with unmanned aerial vehicles.

Google’s Creative Image Search Google Search announces the Creative Image Search feature, enabling users to search for new creative images using various criteria such as color, context, and expression. This opens up a new way to explore and share creative digital art on the internet.

November 2023

Expanding AI Language Translation Capabilities Facebook AI expands its language translation capabilities with a new technology called BabelFish. Unlike traditional language translations, BabelFish uses a multitask LLM model to understand context and generate more natural and accurate translations.

AI Tool for Artists Adobe introduces StudioSketches, a new drawing and creative tool based on artificial intelligence. With the ability to understand artists’ ideas and provide creative support, StudioSketches enhances the workflow of artists and encourages creative freedom.

Machine Learning Platform for Learners Google launches Machine Learning Manor, a new online education platform aimed at creating an authentic learning environment using AI. It combines traditional education with machine learning technology to create a personalized and effective learning experience.

December 2023

Facebook AI Releases Multitask LLM Group Facebook AI announces the Multitask LLM Group model, trained to simultaneously address multiple diverse language tasks. This model is not only effective in solving various tasks but also demonstrates high performance in chaotic and demanding situations that require flexibility.

Elon Musk Announces New Energy Solution Technology Elon Musk announces the New Energy Solution Technology, a Tesla project aiming to create a comprehensive solution for clean energy and global transportation. The project will utilize artificial intelligence and renewable energy sources to minimize emissions and promote the transition to a sustainable energy system.

Drone Artist DJI introduces PhantomCanvas, an autonomous drone capable of creating spatial art. Using machine learning algorithms, it generates complex spatial paintings based on data from sensors and cameras on the unmanned aerial vehicle.