2023’s AI Recap
From progress on the first AI act by the EU to multimodal models like Gemini, from OpenAI’s boardroom turbulence to the rise of industry-specific LLMs, 2023 has been a year of remarkable shifts and advances in AI.
As we look forward to the developments of 2024, let's quickly revisit the key highlights that shaped the AI landscape last year.
1. The introduction of Multimodal AI
Perhaps the most important trend of 2023 was the rise of LLMs and their dominance over AI debates.
CB Insights Research reports a record-high equity funding of nearly $12 billion for Large Language Model (LLM) developers by July 2023. This marks a 12 times increase than the funding raised in 2022.
New models like Inflection-2, Copilot (formerly Bing Chat 2.0), and Claude 2.1 were released, with various companies vying for the top spot as OpenAI’s competitors.
There’s even news of GPT-5 being on the way!
2. The introduction of Multimodal AI
Multimodality is when an AI model can understand various types of input like text, images, audio, video, etc. It allows AI to learn from different sources and understand the context of the information provided.
2023 saw a rise in multimodal AI, with perhaps the most famous being Google’s Gemini. Other multimodal AI models that were launched this year include Meta AI’s ImageBind, LaMDA, and Runway’s Gen-2 model, all of which were released as open-source models.
The appeal of multimodal AI lies in its adaptability, finding applications across industries. In healthcare, for instance, it proves invaluable in aiding physicians in diagnosing complex diseases through the analysis of medical images like MRI, CT scans, and X-rays.
Even ChatGPT, perhaps the most famous generative AI tool, became multimodal in 2023.
3. Greater advancement in AI for photo and video editing
In generative AI, companies are pushing boundaries to infuse heightened realism into visual content. For instance,
Runway's Motion Brush tool, part of its video editing suite, allows precise motion injection into selected image details.
Pika Labs' Pika 1.0 supports 3D animations, cinematic content, and flexible video creation, enabling transformations in background, characters, and minute elements.
Midjourney's V6 image-generating service introduces improved natural language understanding, enhancing its ability to generate more realistic and detailed images.
These companies are continuously working on updates, paving the way for creative enthusiasts to harness AI for unrestricted visual content creation.
4. The rise of Spatial computing
Spatial computing integrates our physical world with virtual experiences, enabling novel interactions between humans and machines.
Humane's AI Pin, unveiled at Paris Fashion Week, transforms your palm into a mini-screen. It facilitates tasks like answering calls through voice commands, summarizing missed emails, and providing nutritional information on food choices.
Nimo Planet revealed its productivity-focused spatial computing system, Nimo Core. The custom rendering system enables resource efficiency, reducing CPU and memory usage. It positions 2D applications in varying depths to render up to six high-fidelity 3D screens within the physical world.
Apart from these, big names like Apple (Vision Pro) and Samsung (XR Glasses) also entered the spatial computing domain with their headsets.
5. AI and Governance
When talking about AI advancements, governance, and regulations are something that cannot be ignored.
The lack of comprehensive AI regulation led to issues like the SAG-AFTRA and WGA strike in the entertainment industry that lasted for 118 days, with AI use being one of the key points of contention.
The Biden administration issued rules for generative AI in October through an executive order. The order outlines eight goals, including establishing new AI safety standards, safeguarding privacy, fostering innovation, and ensuring the government's effective and responsible use of AI.
In December, significant strides were made with the EU AI Act, reaching an agreement with the Council to regulate AI use. Stringent oversight for high-risk AI tools, a ban on harmful practices like indiscriminate image scraping, and restrictions on government use of facial recognition for law enforcement were key components.
6. Advancements in industry-specific LLMs
Closing out this year’s roundup, let's shine a spotlight on how industries like healthcare, law, and finance embraced LLMs to accelerate their productivity. Here are a few that stood out for me:
In healthcare, Google DeepMind introduced Med-PaLM M (MPM), an LLM that uses AI to interpret diverse datasets of clinical notes, lab results, and medical images to assist doctors in diagnosing and treating their patients.
In law, Harvey AI revolutionized the management of legal tasks using GPT-3, offering features like automation and contract analytics, freeing lawyers from administrative tasks.
In finance, BloombergGPT, developed with NVIDIA and Amazon Web Services, employs AI to analyze financial data, making it simpler for financial analysts to extract information.
Beyond these prominent examples, industry-specific LLMs have become prevalent in diverse domains such as food, education, skincare, fashion, entertainment, and more. Each sector now boasts its own flavor of generative AI, tailored to address specific challenges and opportunities.
Looking forward…
Now that you are all caught up, it is time to look forward to further developments in these trends and beyond, preparing ourselves for a year of advancement!
P.S. Which trend are you most excited about in 2024? Stay tuned for my list of top trends to look forward to in 2024!
