Mistral AI, a France-based startup, has introduced a new large language model called Mistral Large, showcasing superior multitask language understanding and excellence in math and coding assessments, though it still hasn't surpassed OpenAI's GPT-4. This is part of a broader AI boom in 2024, with huge gains seen in AI-related cryptocurrencies. Meanwhile, Meta is implementing important measures to safeguard the integrity of the electoral process against AI-generated misinformation ahead of the 2024 European Parliament elections. These efforts include much stricter content labeling and transparency in AI-generated content and advertisements.
France's Answer to the Global AI Race
Mistral AI, a France-based startup, has introduced a new large language model (LLM) named Mistral Large. According to an announcement made on Feb. 26, Mistral Large has demonstrated superior performance in multitask language understanding tests, outshining various well known LLMs, though it still fell short of surpassing OpenAI's GPT-4. The model also excelled in math and coding assessments. However, its capabilities were not compared against recent releases like xAI's Grok and Google's Gemini Ultra.
Guillaume Lample, a founder and chief scientist at Mistral AI, revealed that Mistral Large greatly outperforms the company's earlier models. Alongside Mistral Large, the company launched "Le Chat," an AI chat interface designed to work with its models, similar to how ChatGPT operates on top of GPT-3.5 and GPT-4. Mistral AI, which secured $487 million in funding from major investors including Nvidia, Salesforce, and Andreessen Horowitz in December, boasts that Mistral Large is proficient in over 20,000 English words and fluent in several other languages including French, Spanish, German, and Italian.
Unlike Mistral AI's initial model which was open-source, Mistral Large is proprietary, aligning it with the approach taken by OpenAI for its latest LLMs. This decision has led to some disappointment among the community on X. Despite not being assessed by third-party AI chatbot ranking platforms like Chatbot Arena, Mistral AI's earlier model, Mistral Medium, holds a respectable rank in the AI community.
Mistral AI also announced a partnership with Microsoft, making Mistral Large available on Azure AI Studio and Azure Machine Learning. This collaboration aims to leverage Azure's supercomputing infrastructure for training and scaling Mistral Large.
Pricing for Mistral Large is set at $8 per one million tokens of input and $24 per million tokens of output, positioning it as a slightly more affordable option compared to GPT-4 Turbo.
The AI Crypto Boom
2024 has been very kind to the AI industry so far. The recent announcement of Nvidia's impressive earnings has had a ripple effect across the technology and crypto sectors, particularly boosting the value of artificial intelligence (AI) crypto tokens. After Nvidia's report of a large revenue and earnings increase in Q4 2024, AI-related cryptocurrencies have seen great gains. Worldcoin (WLD), backed by OpenAI CEO Sam Altman, surged by around 240% over the past month. Similarly, Arkham Intelligence's native token, ARKM, experienced a 200+% increase over the same time period.
Nvidia's founder and CEO, Jensen Huang, highlighted the critical juncture for accelerated computing and generative AI, pointing out the worldwide surge in demand across many sectors. This environment has provided a very favorable backdrop for crypto projects integrating AI technologies. For instance, Sleepless, a Web3 AI companion platform supported by Binance Ventures, unveiled its 2024 roadmap, which includes the launch of its AI blockchain game, Her 2.0. The game promises authentic emotional communication through AI interactions, alongside gameplay features that allow dressing virtual partners, engaging in battles, and earning rewards.
In addition, Fetch.ai announced a partnership with Deutsche Telekom to incorporate enterprise AI solutions, marking a big step for the AI platform known for its autonomous agents facilitating transactions and resource management across industries. This collaboration is set to expand Fetch.ai's reach and utility, particularly with Deutsche Telekom acting as a validator on the Fetch.ai blockchain.
The excitement around AI in the crypto space is further evidenced by the increase in trading volumes and participant numbers for AI tokens on the crypto exchange Bitget. This surge could certainly be attributed to the market's reaction to OpenAI's launch of its text-to-video model Sora.
Moreover, the decentralized computing blockchain iExecRLC, which has been exploring AI edge computing since 2020 as part of Nvidia’s Inception Program, now boasts a market capitalization of $323 million. Y Combinator's encouragement for startups to develop solutions in stablecoin, metaverse, and AI also reflects a broader industry trend towards the normalization of digital innovations.
Meta's Strategy to Safeguard Elections
Unfortunately, not all the news about AI has been positive. Meta, the tech giant behind Facebook and Instagram, is taking important steps to safeguard the integrity of the electoral process on its platforms ahead of the 2024 European Parliament elections.
In a recent blog post, Marco Pancini, Meta's head of EU Affairs, detailed the company's strategy to combat the misuse of generative artificial intelligence (AI). This includes applying the principles of Meta’s "Community Standards" and "Ad Standards" to AI-generated content, which will also be subject to review by independent fact-checking partners to identify if the content has been altered or manipulated.
The company also emphasized that photorealistic images generated using Meta’s own AI tools must be labeled accordingly. Extending these measures, Meta announced plans to introduce features that will allow users to label AI-generated content created by external tools from companies like Google, OpenAI, Microsoft, Adobe, Midjourney, and Shutterstock. This should enhance transparency, especially when users share AI-generated videos or audios, with potential penalties for those failing to disclose if they do so.
Meta is also tightening its grip on political, social, and election-related advertisements that use AI. Advertisers have to mention if their content was altered or created using AI technologies. Meta already reported that it removed 430,000 ads across the European Union between July and December of 2023 for lacking the necessary disclaimers.
This initiative by Meta is part of a broader industry effort to address the challenges posed by AI in the context of major global elections in 2024. Other tech giants, including Google and OpenAI, have also implemented or announced policies to mitigate the risks of AI-generated political misinformation. Google, for instance, plans to limit responses to election queries on its AI chatbot to prevent the spread of false information during the U.S. presidential election.
Furthermore, 20 companies, including industry leaders like Microsoft, Google, Anthropic, Meta, OpenAI, Stability AI, and X, have pledged to combat AI election interference.
The global response extends beyond the tech industry, with governments and regulatory bodies also taking steps to counter the misuse of AI. The European Commission, for example, has initiated public consultations on election security guidelines to combat the democratic threats posed by generative AI and deepfakes. In the U.S, AI-generated voices in automated phone scams have been outlawed after incidents of deepfake scams.