2023 shall be remembered as the year in which AI “arrived.” While we’ve been seeing AI make rapid progress and be increasingly impactful for over a decade now, this past year was the first time it has been directly used and discussed by a large portion of humanity. Unsurprisingly, that in turn led to much more media coverage of AI than ever before; in 2023 alone, Last Week in AI curated over 3000 articles about AI across 52 newsletter releases.
In this piece, we’ll take a look back over the stories we highlighted throughout the year.
January
Highlights:
AI Is Now Essential National Infrastructure
A Professional Artist Spent 100 Hours Illustrating This Book Cover, Only To Be Accused Of Using AI
A.I. Turns Its Artistry to Creating New Human Proteins
Microsoft’s new AI can simulate anyone’s voice with 3 seconds of audio
Inside Japan’s long experiment in automating elder care
Anthropic’s Claude improves on ChatGPT but still suffers from limitations
Conservatives Are Panicking About AI Bias, Think ChatGPT Has Gone ‘Woke’
Deepfake challenges ‘will only grow’
Google created an AI that can generate music from text descriptions, but won’t release it
BuzzFeed says it will use AI tools from OpenAI to personalize its content
The best use for AI eye contact tech is making movie stars look straight at the camera
Newsletter links: Week 1 | Week 2 | Week 3 | Week 4
Our Thoughts: ChatGPT was first released in late 2022 and became an unexpected viral sensation, so in January 2023 the impact of its release was still being felt through the many discussions regarding AI’s significance. Claude’s release marked the first of many competing chatbots to come out throughout 2023, though none managed to dethrone OpenAI’s tech. BuzzFeed became the first of multiple media groups to announce plans to adopt AI for content generation. Overall, January kicked the year off with ChatGPT and its ilk being the main theme of AI news, and it has remained that way since.
February
Highlights:
Whispers of A.I.’s Modular Future
The current legal cases against generative AI are just the beginning
OpenAI releases tool to detect AI-generated text, including from ChatGPT
A Tech Race Begins as Microsoft Adds A.I. to Its Search Engine
Meta, Long an A.I. Leader, Tries Not to Be Left Out of the Boom
Bing’s A.I. Chat: ‘I Want to Be Alive. 😈’
ChatGPT AI passes test designed to show theory of mind in children
Audiobook Narrators Fear Spotify Used Their Voices to Train AI
How I Broke Into a Bank Account With an AI-Generated Voice
Introducing LLaMA: A foundational, 65-billion-parameter large language model
AI-Human Romances Are Flourishing—And This Is Just the Beginning
Newsletter links: Week 1 | Week 2 | Week 3 | Week 4
Our Thoughts: Microsoft’s release of AI updates to its Bing browser brought about the second ChatGPT-esque chatbot that anyone could try, and with it a general sense that Microsoft is now leading the pack while Google is lagging in commercializing AI tech. Appropriately, Google hurried to announce the planned release of its chatbot, Bard. Meta emphasized to investors it is working to build AI into its products as well and also announced what would be the first of many powerful open-source AI models with LLaMA. At the same time, lawsuits and debates concerning the legality of generative AI, and its impact on the jobs of people in various creative industries, began to emerge.
March
Highlights:
Will artificial intelligence replace your lawyer–and will its name be Harvey?
Meta says it is experimenting with AI-powered chat on WhatsApp and Messenger
Meta’s powerful AI language model has leaked online — what happens now?
Alpaca: A Strong, Replicable Instruction-Following Model
Microsoft and Google Unveil A.I. Tools for Businesses
Google Releases Bard, Its Competitor in the Race to Create A.I. Chatbots
People Aren’t Falling for AI Trump Photos (Yet)
Newsletter links: Week 1 | Week 2 | Week 3 | Week 4
Our Thoughts: March saw the continuation of trends that kicked off in February. Multiple new commercial chatbots came out (Google released Bard, OpenAI’s upgraded ChatGPT with GPT-4, and Anthrophic’s Claude became more widely available), as did several open-source ones (Meta’s LLaMA was leaked and became downloadable by anyone despite being meant for academic use, and Stanford had released Alpaca, the first of many ‘descendants’ of LLaMA). Beyond that, both Microsoft and Google kicked off the effort to integrate AI across their entire range of products, an effort they will continue for the rest of the year.
April
Highlights:
Clearview AI used nearly 1m times by US police, it tells the BBC
Microsoft’s Bing chatbot is getting ads
Midjourney ends free trials of its AI image generator due to ‘extraordinary’ abuse
Canada Opens Probe into OpenAI, the Creator of AI Chatbot ChatGPT
Measuring trends in Artificial Intelligence
OpenAssistant RELEASED! The world’s best open-source Chat AI!
Google Devising Radical Search Changes to Beat Back A.I. Rivals
Align your Latents: High-Resolution Video Synthesis with Latent Diffusion Models
Stack Overflow Will Charge AI Giants for Training Data
Google to combine AI research units Google Brain, DeepMind
Newsletter links: Week 1 | Week 2 | Week 3 | Week 4
Our Thoughts: April was relatively quiet, with developments to stories we’ve seen introduced in months prior but no new major developments. One exception is Stack Overflow’s announcement of charging for access to its data for training large models, which is a percentage multiple others followed soon after.
May
Highlights:
‘The Godfather of A.I.’ Leaves Google and Warns of Danger Ahead
Inside the Discord Where Thousands of Rogue Producers Are Making AI Music
No Cloud Required: Chatbot Runs Locally on iPhones, Old PCs
Meta unveils A.I. ‘testing playground’ to help advertisers build campaigns
U.S. Sanctions Drive Chinese Firms to Advance AI Without Latest Chips
EU AI Act To Target US Open Source Software
Why use of AI is a major sticking point in the ongoing writers’ strike
Newsletter links: Week 1 | Week 2 | Week 3 | Week 4 | Week 5
Our Thoughts: Hinton’s departure from Google and move to warning about AI risk was a major development in the conversation regarding the potential of AI to harm humanity. Production of music with AI really started taking off around this time, just as Hollywood writers were on strike partially because of AI and its potential uses in film and TV production.
June
Newsletter links: Week 1 | Week 2 | Week 3 | Week 4
Highlights:
Europe approves landmark AI legislation, challenging tech giants’ power
Paul McCartney says AI tools helped rescue John Lennon vocals for ‘last Beatles record’
Meta announces Voicebox, a generative model for multiple voice synthesis tasks
China’s ByteDance Has Gobbled Up $1 Billion of Nvidia GPUs for AI This Year
State Department considers generative AI for contract writing
OpenLLaMA is a fully open-source LLM, now ready for business
We should all be worried about AI infiltrating crowdsourced work
Our Thoughts: June continues the steady upward trend of public LLM releases throughout the year. OpenLLaMA was an especially important development as one of the first truly open-source LLMs (open-source the model code, model weights, and training data) that could be used for commercial purposes.
July
Newsletter links: Week 1 | Week 2 | Week 3 | Week 4
Highlights:
ChatGPT maker OpenAI faces class action over how it used people’s data
‘It’s not like science fiction any more’: Nasa aiming to make spaceships talk
OpenAI Plans ChatGPT ‘Personal Assistant for Work,’ Setting Up Microsoft Rivalry
How the Great AI Flood Could Kill the Internet
NYC’s anti-bias law for hiring algorithms goes into effect
EU and Japan look to partner on A.I. and chips as China ‘de-risking’ strategy continues
OpenAI’s ChatGPT Shows a Significant Slowing in Traffic
Google’s medical AI chatbot is already being tested in hospitals
Elon Musk unveils his AI company, X.AI
65% Of Top AI Companies Have Immigrant Founders
Aided by A.I. Language Models, Google’s Robots Are Getting Smart
An AI Startup Is Helping North American Diesel Trains Clean Up Their Act
NYC subway using AI to track fare evasion
Meta and Microsoft Introduce the Next Generation of Llama
Our Thoughts: The large model release frenzy is picking up steam, with Anthropic’s Claude 2 (an impressive 100k context length, but only API access) and Meta’s LLaMA 2 (open source model and weights, and it can be used commercially). Meanwhile, we’re seeing more copyright lawsuits develop against generative AI.
August
Newsletter links: Week 1 | Week 2 | Week 3 | Week 4
Highlights:
Mind Over Paralysis: AI Helps Quadriplegic Man Move and Feel Again
Massachusetts regulators launch probe into AI in securities industry
Movie extras worry they’ll be replaced by AI. Hollywood is already doing body scans
Meta is reportedly preparing to release AI-powered chatbots with different personas
Eight Months Pregnant and Arrested After False Facial Recognition Match
White House launches AI-based contest to secure government systems from hacks
Spotify expands its AI-powered DJ feature globally
Bots are better at beating ‘are you a robot?’ tests than humans are
The New York Times prohibits AI vendors from devouring its content
FEC could limit AI in political ads ahead of 2024
As Hollywood Strikes, 96% Of Entertainment Companies Are Boosting Generative AI Spend
How Nvidia Built a Competitive Moat Around A.I. Chips
Introducing Code Llama, an AI Tool for Coding
Reinforced Self-Training (ReST) for Language Modeling
Our Thoughts: August saw a mix of developments with advances in both models and regulations. Garner placed generative AI at the “peak of inflated expectations” for the 2023 hype cycle. Looking back now, this may not have been correct – generative AI had a lot of hype back in August, but it still has a lot of hype now, arguably more so. It’s probably still not close to the “trough of disillusionment,” the next step in the hype cycle.
September
Newsletter links: Week 1 | Week 2 | Week 3 | Week 4
Highlights:
AI21 Labs Raises $155M in Series C Funding; Valued at $1.4 Billion
Google’s Duet AI is now available in Gmail, Docs, and more for $30 a month
OpenAI launches a ChatGPT plan for enterprise customers
It’s a Weird Time for Driverless Cars
Pentagon plans vast AI fleet to counter China, Wall Street Journal reports
Imbue raises $200M to build AI models that can ‘robustly reason’
Ads for AI sex workers are flooding Instagram and TikTok
Anthropic’s Claude AI chatbot gets a paid plan for heavy users
Salesforce introduces new AI assistant, Einstein Copilot, for all its CRM apps
California bill to ban driverless autonomous trucks goes to Newsom’s desk
Adobe starts paying bonuses to Stock contributors whose content is being used to train Firefly
DeepMind discovers that AI large language models can optimize their own prompts
Paper: LLMs trained on “A is B” fail to learn “B is A”
Teachers Are Going All In on Generative AI
Our Thoughts: September’s news made clear, if it wasn’t already, that incumbent tech companies, which already have wide-reach products and distribution channels, would benefit immediately from generative AI. Product releases from Microsoft, Google, Adobe, and Salesforce show how easy it is to integrate generative AI into existing software, and how hard it will be for upstarts to disrupt existing product categories with generative AI alone.
October
Newsletter links: Week 1 | Week 2 | Week 3 | Week 4 | Week 5
Highlights:
Mistral AI makes its first large language model free for everyone
Hollywood’s Deal With Screenwriters Just Rewrote the Rules Around A.I.
Franzen, Grisham and Other Prominent Authors Sue OpenAI
Top GOP senator teams up with key Dem on ‘light-touch’ AI bill
Decomposing Language Models Into Understandable Components
Scaling GAIA-1: 9-billion parameter generative world model for autonomous driving
Introducing Stable LM 3B: Bringing Sustainable, High-Performance Language Models to Smart Devices
‘A.I. Obama’ and Fake Newscasters: How A.I. Audio Is Swarming TikTok
Adobe is upgrading Photoshop’s generative AI model — and releasing more for Illustrator and Express
Humanoid robots face a major test with Amazon’s Digit pilots
Adept Releases Fuyu-8B for Multimodal AI Agents
4K4D: Real-Time 4D View Synthesis at 4K Resolution
White House to unveil sweeping AI executive order next week, tackling immigration, safety
Cruise Self-Driving License Revoked After It Withheld Pedestrian Injury Footage, DMV Says
This new data poisoning tool lets artists fight back against generative AI
Our Thoughts: Many generative AI models were again released in October, but we’ll briefly turn our attention to the world of self-driving cars, where after a gruesome accident, Cruise’s license was revoked by the DMV. In the weeks that followed, Cruise saw massive personnel changes, from the departures of key executives, to layoffs, and funding cuts from GM. This leaves Waymo as the only company with commercially available self-driving taxis at the moment, as well as perhaps the only such company in the near future.
November
Newsletter links: Week 1 | Week 2 | Week 3 | Week 4
Highlights:
Biden Issues Executive Order to Create A.I. Safeguards
ChatGPT is combining its different abilities into a single ‘Voltron-style’ chat
Elon Musk debuts ‘Grok’ AI bot to rival ChatGPT, others
At UK’s AI Summit developers and govts agree on testing to help manage risks
Here’s what we know about generative AI’s impact on white-collar work
Musk’s xAI set to launch first AI model to select group
EU’s AI Act negotiations hit the brakes over foundation models
Humane’s AI Pin: all the news about the new AI-powered wearable
Sam Altman fired as CEO of OpenAI
Google DeepMind wants to define what counts as artificial general intelligence
Lawsuit claims UnitedHealth AI wrongfully denies elderly extended care
Five Days of Chaos: How Sam Altman Returned to OpenAI
Cruise co-founder and CEO Kyle Vogt resigns
Stability AI debuts Stable Video Diffusion models in research preview
Underage Workers Are Training AI
Our Thoughts: The craziest boardroom drama (at least for AI companies) came and went in November. For five days, Sam Altman was ousted as OpenAI’s CEO. Through a series of increasingly surprising turn of events (OpenAI employees threaten to resign en masse if Altman doesn’t return, Microsoft promises to hire all OpenAI employees at equal compensation, OpenAI board reaching out to Anthropic for a potential merger, former CEO of Twitch hired as OpenAI CEO for a couple of days, and the list goes on…), Altman finally returned to OpenAI, with a new board, and with Microsoft serving an observing role on that board. Definitely the most memorable AI news of this past year.
December
Newsletter links: Week 1 | Week 2 | Week 3 | Week 4
Highlights:
ChatGPT’s 1-year anniversary: how it changed the world
Sam Altman returns as CEO, OpenAI has a new initial board
Amazon Introduces Q, an A.I. Chatbot for Companies
Making an image with generative AI uses as much energy as charging your phone
Alphabet unveils long-awaited Gemini AI model
E.U. Agrees on Landmark Artificial Intelligence Rules
OpenAI Rival Mistral Nears $2 Billion Valuation With Andreessen Horowitz Backing
Phi-2: The surprising power of small language models
Mamba: Linear-Time Sequence Modeling with Selective State Spaces
Deepfakes for $24 a month: how AI is disrupting Bangladesh’s election
AI image training dataset found to include child sexual abuse imagery
Anthropic will help users if they get sued for copyright infringement.
Midjourney V6 is here with in-image text and completely overhauled prompting
Our Thoughts: It’s been one year since ChatGPT’s release. Despite all the hype, the world has changed. Generative AI models and applications have developed and proliferated far beyond what the initial ChatGPT was capable of. Insane amounts of money and human resources are being poured into this sector. AI has matured well into the mainstream digest, and we maintain the characterization that 2023 was still just the beginning of AI’s transformative impact. We expect 2024 will see faster progress, as more organizations adopt AI and as AI costs shrink. How powerful will GPT-5 be? How will governments regulate copyrights for generative AI? Will the AI hype surge higher or fizzle out? We are excited to find out.
Read More in Last Week in AI
