The Air Street Capital report pointed out that large-scale language models continue to make breakthroughs in various fields, computing power has become a new competitive advantage, and generative AI applications have received a large amount of investment. In addition, the report also predicts ten major trends in the future, including generative AI for blockbuster production, self-improving AI agents, etc.
Air Street Capital, a venture capital firm specializing in the field of AI, recently released the “2023 Artificial Intelligence Status Report.” This is the sixth consecutive year that the venture capital firm has released the report. This year’s more than 150-page report covers artificial intelligence research, industry, security and politics, and predicts industry trends in 2024. Air Street Capital’s predictions last year were correct about 50% of the time.
Air Street Capital’s report last year outlined the rise of decentralization in artificial intelligence research, but OpenAI’s GPT-4 shocked observers as big tech companies made a comeback. In the scramble for more computing power, challengers have found themselves increasingly dependent on its cash reserves. Meanwhile, the open source community continues to thrive as the number of language models released continues to surge.
Air Street Capital pointed out in the report that currently published technical reports on the most advanced large-scale language models do not contain useful information for artificial intelligence researchers and are increasingly “closed source.” One of OpenAI’s co-founders even described their original open source philosophy as “completely wrong.” In contrast, open source Meta AI has emerged as a champion of open artificial intelligence, with their family of LLaMa models currently the most powerful publicly available alternative.
Key points of the “2023 Artificial Intelligence Status Report”:
●OpenAI’s GPT-4 remains the king of large language models, beating all other large language models on both classic benchmarks and exams designed to evaluate humans.
● Increasingly, attempts are being made to clone or surpass proprietary performance through smaller models, better data sets, and longer contexts. These may gain new urgency amid concerns that human-generated data may dry up, only sustaining AI trends in the coming years.
●Large-scale language models and diffusion models continue to drive real-world breakthroughs, particularly in the life sciences, with meaningful progress in both molecular biology and drug discovery.
●Computing power is like the new oil. Nvidia’s revenue hit a record high, and startups waved their Nvidia GPUs as a competitive advantage.
GenAI saves the VC world, with AI startups focused on generative AI applications including video, text and coding raising more than $18 billion from VCs and corporate investors at a time when tech company valuations are plummeting USD funds.
●The safety debate has become mainstream, prompting governments and regulators around the world to take action. However, this flurry of activity masks divisions within the AI community and a lack of concrete progress in global governance, as governments around the world take conflicting approaches.
Evaluating state-of-the-art models is increasingly challenging because standard large language models often struggle with robustness. Consider the risks, as the “vibes-based” approach is not good enough.
Top 10 predictions for 2024:
①Create a Hollywood-style blockbuster using generative AI for visual effects.
②A generative AI media company is under investigation for abuse during the 2024 US election.
③Self-improving AI agents crush SOTA in complex environments (e.g. AAA games, tool usage, science).
④The tech IPO market is starting to loosen up, with at least one artificial intelligence-focused company, such as Databricks, going public.
⑤The generative AI scaling boom resulted in one team spending more than $1 billion to train a single large model.
⑥The US FTC or the UK CMA investigate the Microsoft/OpenAI transaction on the grounds of monopoly.
⑦Beyond high-level voluntary commitments, progress in global AI governance has been limited.
⑧Financial institutions launch GPU debt funds to replace equity investment in venture capital for financing.
⑨An AI-generated song cracks the Billboard top 10 or Spotify’s 2024 Hot Songs chart.
⑩ As the workload and cost of reasoning have increased significantly, large artificial intelligence companies (such as OpenAI) have acquired an artificial intelligence chip company that focuses on reasoning.
Source of this article: Tencent Technology, original title: “2023 Artificial Intelligence Status Report: Computing power is like the new oil, generative AI saves venture capital”
Risk warning and disclaimer
Market risk, the investment need to be cautious. This article does not constitute personal investment advice, nor does it take into account the special investment objectives, financial situation or needs of individual users. Users should consider whether any opinions, views or conclusions contained in this article are appropriate to their particular circumstances. Invest accordingly and do so at your own risk.