March 2024 marked the one-year anniversary of the release of OpenAI’s GPT-4, the most powerful AI language model to date. The excitement surrounding generative AI shows no signs of slowing down, with major tech companies from around the world making significant announcements in the past month.
Key Developments in March
- Global AI Chip Race Heats Up: The release of the B200, the world’s most powerful AI chip, has sparked a new wave of competition in the AI chip market.
- OpenAI Founder Sam Altman’s Net Worth Surpasses $14 Billion: Altman’s success with OpenAI has made him one of the wealthiest people in the world.
- Generative AI Market to Reach $1.3 Trillion by 2032: A new report from Bloomberg AI Research predicts that the generative AI market will grow at a CAGR of 43% over the next eight years.
This article provides a comprehensive overview of the major news events in the general artificial intelligence (AGI) industry in March 2024. It aims to help readers stay up-to-date on the latest developments in this rapidly evolving field and understand the transformative potential of AI.
In the following sections, we will explore each of these developments in more detail.
March 2
- Google admits flaws in Gemini’s image generation: Google co-founder Sergey Brin admitted at a hackathon focused on Google’s large language model Gemini 1.5 that Gemini’s image generation is “messed up” and apologized for it. He said that Google did not conduct sufficient testing before launching Gemini, which was the key reason for the problem.
March 4
- OpenAI founder Altman’s net worth revealed: According to the latest Bloomberg Billionaires Index, Sam Altman, the 38-year-old founder and CEO of OpenAI, has a net worth of at least 2 billion USD, but this figure does not include any of his shares in OpenAI. Altman has repeatedly stated that he does not own any shares in OpenAI. Most of his wealth is held in venture capital firms and startup investment projects, including the Reddit IPO.
- Alibaba leads funding round for MiniMax: Alibaba leads a $600 million+ financing round for general large language model startup MiniMax, becoming the core lead investor in the round. MiniMax has previously completed three rounds of financing, with investors including Tencent Investment.
- Anthropic releases Claude 3 series: Anthropic, a competitor to OpenAI backed by Google and Amazon, releases the latest Claude 3 series of models, including Claude 3 Opus, Sonnet, and Haiku, with the strongest model Opus outperforming OpenAI’s GPT-4.
March 5
- US blocks AMD’s “China-specific” AI chip sale: The US government blocks AMD from selling its “China-specific” AI chip MI309 to Chinese customers. The chip is reportedly called MI309 and is designed to meet new US export control regulations. However, the US government currently does not approve the export of this special chip to China. US officials told AMD that the MI309 AI chip is still too powerful and exceeds the restriction requirements, and the company must obtain a license from the Bureau of Industry and Security (BIS) to sell the chip.
March 6
- Perplexity’s valuation to reach $1 billion: AI startup Perplexity, which aims to challenge Google’s dominance in web search, is set to close a new round of financing, with the company’s valuation expected to reach nearly 1 billion USD, about double its latest valuation round a few months ago.
March 8
- Italy investigates OpenAI’s Sora: Italian regulator Garante asks OpenAI to clarify whether the way it informs users and non-users about the data used in its product Sora complies with EU regulations.
March 9
- Generative AI market predicted to reach $1.3 trillion: A new report by Bloomberg predicts that the generative AI market will grow to $1.3 trillion by 2032, with a compound annual growth rate of 43%.
- OpenAI CEO Altman reinstated: OpenAI, a US company, tweeted that after months of investigation, the final conclusion on the dismissal of OpenAI co-founder and CEO Altman was that Altman did not engage in any misconduct last year, and the board of directors fired him for no reason. Therefore, Altman will continue to serve as the company’s CEO and director.
March 10
- ChatGPT’s massive electricity consumption: According to reports, OpenAI’s popular chatbot ChatGPT may consume over 500,000 kilowatt-hours of electricity per day to respond to about 200 million user requests. In comparison, the average American household uses about 29 kilowatt-hours of electricity per day, which means that ChatGPT consumes more than 17,000 times as much electricity as a household. If generative AI is further adopted, the power consumption is likely to be even higher.
March 13
- EU passes landmark AI Act: The European Parliament passed the landmark Artificial Intelligence Act by a vote of 523 to 46, with 49 abstentions. The law will ban certain AI applications that “threaten citizen rights”, including biometric classification systems based on sensitive characteristics, and the indiscriminate scraping of facial images from the internet or CCTV footage to create facial recognition databases. AI that manipulates human behavior or exploits human weaknesses will also be banned.
- World’s first AI programmer Devin launched: Startup Cognition AI launches the world’s first AI programmer Devin, whose biggest breakthrough is that it greatly improves computer reasoning and planning capabilities. It is understood that Devin has mastered multiple capabilities, including self-learning new technologies, building and deploying applications, autonomously finding and fixing bugs, and training and fine-tuning his own AI models.
- Figure releases robot powered by OpenAI models: Robotics startup Figure releases a demonstration of its first robot powered by OpenAI’s large language models, Figure 01. Despite using only one neural network, the robot can follow human commands, such as handing them apples, picking up black plastic bags, and putting cups and plates away.
March 14
- OpenAI says Sora will be released to the public later this year: OpenAI’s CTO Mira Murati said recently that Sora will be officially released to the public later this year, “probably in a few months”. OpenAI will support sound effects in Sora and allow users to edit the video content generated by Sora.
- Together AI valued at $1.25 billion: Nvidia-backed generative AI startup Together AI has raised $106 million in a new round of funding at a valuation of $1.25 billion, led by Salesforce Ventures. Together AI, founded in 2022, is an open-source platform that developers can use to build or customize their own AI models.
- OpenAI to launch French and Spanish ChatGPT versions: OpenAI announced that it has signed licensing agreements with two major European publishers, French newspaper Le Monde and Spanish group Promotora de Informaciones SA or Prisa, to bring French and Spanish news content into ChatGPT and help train OpenAI models.
March 16
- Apple reportedly acquires Canadian startup DarwinAI: Apple acquired Canadian AI startup DarwinAI earlier this year, and dozens of DarwinAI employees have joined Apple’s AI division.
- Apple launches 300-billion parameter MM1 multimodal large model: In a paper titled “MM1: Methods, Analysis & Insights from Multimodal LLM Pre-training” authored by multiple authors, Apple officially announced its research results on multimodal large models. This is a series of multimodal LLMs with up to 300 billion parameters, which consists of dense models and Mixture of Experts (MoE) variants. It not only achieves the best level in pre-training metrics, but also maintains competitive performance after supervised fine-tuning on a series of existing multimodal benchmarks.
March 18
- Nvidia Unveils Most Powerful AI Chip Yet, the B200: Nvidia CEO Jensen Huang announced the new generation of GPUs, Blackwell, at the Nvidia GTC conference in San Jose, California. The first generation Blackwell chip, codenamed GB200, will be released later this year.
- Blackwell has 208 billion transistors and is manufactured using TSMC’s 4nm process. The previous generation GPU, Hopper H100, used a 4nm process and had 800 billion transistors.
- Elon Musk’s xAI Grok-1 Model Goes Open Source: Elon Musk’s AI startup xAI announced that its large language model, Grok-1, is now open source. Users can directly download the model’s basic weights and network architecture information via a magnet link.
March 20
- OpenAI Plans to Release GPT-5 in Mid-2024: According to anonymous sources, OpenAI is planning to release GPT-5 in mid-2024, with a target release date of summer.
March 21
- Stable Diffusion Core Team Reportedly Resigns: The core research team of Stable Diffusion has reportedly resigned collectively. The list includes the research team leader and first author of the paper, Robin Rombach, co-author Andreas Blattmann, and another author Dominik Lorenz. According to an informed source, Stability AI is in trouble due to financial difficulties and inability to raise new funds.
- Moon’s Dark Side Kimi Intelligent Assistant Crashes: After testing, it was found that the APP and mini program of the large model application Kimi under Moon’s Dark Side cannot be used normally. Previously, Moon’s Dark Side issued a statement: Starting from 9:30 am on March 20th, it was observed that Kimi’s system traffic continued to increase abnormally, and the trend of traffic increase far exceeded its resource planning expectations. This led to a large number of SaaS customers experiencing the abnormal problem of “429:engine is overloaded” continuously from 10:00 am on the same day.
March 22
- Suno v3 Model Generates Complete Songs in Seconds: AI music creation startup Suno AI has launched its AI music generation model, Suno v3. This is the first model capable of generating radio-quality music, and can create complete 2-minute songs in seconds. v4 is already under development.
- Baidu to Provide AI Features for Apple’s Latest Domestic Smartphones: According to informed sources, Baidu will provide AI features for Apple’s iPhone 16, Mac system, and iOS 18 to be released this year. Apple had previously negotiated with Alibaba and another domestic large model company, and finally decided to choose Baidu to provide this service. Apple is expected to charge through the API interface method.
March 25
- HeyGen Raises $60 Million in New Funding Round, Valuation Increases Sixfold in Four Months: HeyGen, an AI video platform, is raising $60 million in a new round of funding, valuing the company at $440 million pre-money, a sixfold increase from its valuation just four months ago. The round is being led by Benchmark, an early investor in companies like Uber.
March 26
- Canada to Strengthen Foreign Investment Review in Areas Like AI: Canada’s industry minister said on Monday that under revised rules to the Investment Canada Act, non-Canadian companies will be required to notify the government before investing in or acquiring Canadian entities in key technology sectors such as AI and quantum computing. The move is seen as a follow-up to similar measures taken by the United States in an effort to curb Chinese capital acquisitions.
March 27
- OpenAI poised to partner with Hollywood to revolutionize the film industry: OpenAI has been holding a series of promotional events in Hollywood recently, hoping to showcase the enormous potential of its latest AI text-to-video generative model, called Sora, to film studios, talent agencies, and executives. Some analysts believe that this could be the “most significant industry meeting in Hollywood history” and could have a groundbreaking impact on the development of the global film industry.
- 1320 billion open-source large model DBRX goes online: Databricks has launched a general-purpose large language model called DBRX, which it claims is the “most powerful open-source AI to date” and has reportedly outperformed “all open-source models on the market” in various benchmarks.
March 28
- Amazon Invests an Additional $2.75 Billion in Anthropic: Amazon has made an additional $2.75 billion investment in AI startup Anthropic, bringing its total investment to $4 billion. This is Amazon’s largest venture capital investment to date. The investment follows a $1.25 billion investment made in December 2023.
- Google and Stanford University Launch AI Verification Tool: Google DeepMind’s research team has published a new study introducing an AI system called “Search-Augmented Factuality Evaluator (SAFE)”. The system uses a large language model to break down generated text into individual facts and then uses Google Search results to determine the accuracy of each statement.
March 30
- Microsoft and OpenAI to Invest $100 Billion in Building AI Supercomputer: Microsoft and OpenAI are planning to invest $100 billion to build an AI supercomputer called “Stargate”. OpenAI’s next major AI upgrade is expected to be released early next year. Microsoft executives hope to release the “Stargate” AI supercomputer by 2028 at the earliest. In addition, Microsoft and OpenAI plan to develop a data center project for AI.
- OpenAI Releases Voice Engine, a Speech Model That Can Replicate Original Audio in 15 Seconds: OpenAI has unveiled its new custom audio model, “Voice Engine”, on its official website. With just 15 seconds of reference audio, Voice Engine can generate new audio that is almost identical to the original in terms of clarity, fluency, timbre, and naturalness. It is significantly better than most products on the market.
March 31
- U.S. House Bans Microsoft AI Copilot: The U.S. House of Representatives has implemented a strict ban on the use of Microsoft’s AI chatbot Copilot by congressional staffers. The Office of Cybersecurity cited the Copilot application as a risk to users due to the potential for it to leak House data to non-House-approved cloud services.
- Chinese Team Develops “Human Face Robot” that Imitates Human Expressions by Looking in the Mirror: A research team of Chinese engineers at Columbia University’s Engineering School recently released an innovative robot called Emo. This robot has the ability to predict and mimic human facial expressions, and can also make eye contact. Emo’s unique feature is its self-supervised learning framework, which operates similarly to how humans practice facial expressions by looking in the mirror.
