Open-Source AI Models Expected to Outpace Centralized Counterparts by 2024, ARK Invest Predicts
Summary:
Recent research from ARK Invest suggests open-source AI models are closing the performance gap with centralized AI models led by Google and OpenAI. The data indicates that open-source AIs could surpass their centralized counterparts by 2024. Top-performing open-source models in 2023, such as Yi 34B, Falcon 180B, and Mixtral 8x7B, showcased performance comparable to market leaders. The study underscores the importance of democratizing access to generative AI.
The leading performance in the realm of generative artificial intelligence (AI) models currently comes from centralized cloud infrastructure-backed machines, such as ChatGPT. However, recent studies indicate that open-source competitors might be closing the gap. The forerunners of generative AI today, including Google and OpenAI, consolidated their infrastructure, effectively resulting in limited public access to certain information, such as the training model's data sources. This scenario could soon change, according to ARK Invest's research team, led by Cathy Wood. Their data suggest that by 2024, open-source AI models might surpass centralized models in performance.
[Image: Comparison of open-source and proprietary AI models performance. Source: ARK Invest]
The image above portrays the advancements achieved by open-source AI models since 2022. Interestingly, some of these models have outperformed proprietary ones. The AI landscape is currently ruled by companies like OpenAI, Google, and its parent firm, Alphabet. These giants offer popular models like ChatGPT-4 and Gemini Ultra. On the contrast, companies like Meta (formerly Facebook), Mistral, and a few Chinese companies adopted an open-source strategy.
In 2023, a few top-performing open-source AI models surfaced, such as Yi 34B, Falcon 180B, and Mixtral 8x7B. These models demonstrated a performance on par with the market leaders. Jozef Soja, an ARK Invest researcher, pointed out that Mixtral surpassed GPT 3.5 in performance on the Massive Multitask Language Understanding (MMLU) benchmark, which underlines the current lead of GPT-4. Parallelly, Meta's foundational model, LLaMA, showed significant improvements in its 2023 versions. The necessity of such progress in open-source AI models emerges from the requirement to democratize access to generative AI, as explained by AI engineer Brian Roemmele.
The current market leaders in AI — Google's Gemini and OpenAI's ChatGPT-4 — have been recently evaluated. Cointelegraph posed several cryptocurrency-related questions to the free versions of Gemini (through Bard) and ChatGPT-4 and compared their responses. Both AIs cautioned users seeking investment advice and urged them to consult professional financial advisors for more personalized counsel. Despite having some variance in the details, both models made similar considerations in their responses.
Published At
12/14/2023 11:48:09 AM
Disclaimer: Algoine does not endorse any content or product on this page. Readers should conduct their own research before taking any actions related to the asset, company, or any information in this article and assume full responsibility for their decisions. This article should not be considered as investment advice. Our news is prepared with AI support.
Do you suspect this content may be misleading, incomplete, or inappropriate in any way, requiring modification or removal?
We appreciate your report.