| | | | THE WEEKLY NEWSLETTER OF AIM Monday, 29 jubne, 2024 | Was this email forwarded to you? Sign up here By Amit Raja Naik, Editor AT AIM | | All in all, the AI race has intensified, promising thrilling advancements ahead. Building on the momentum of last week's major AI announcements, let’s delve deeper into the implications and future directions of this technology. | | | | | | | The Linux Moment in AI Arrives | | | | A few days ago, AIM asked a pertinent question: “When will the ‘Linux moment’ in AI arrive?” Looks like it has arrived sooner than expected. Meta released Llama 3.1, outperforming OpenAI’s GPT-4o on several benchmarks, making it the best open-source model available to date. It has scaled all the way to the top of the leaderboard. Chief Mark Zuckerberg is confident that Llama 3.1 will have a similar impact on the AI ecosystem as Linux had on the operating system world. Read the full story here. | | | | | | | | | | | | | | As the industry buzzes about Meta's latest Llama 3.1 405B release, Groq announced its proud partnership with Meta, making the latest Llama 3.1 models—including 405B Instruct, 70B Instruct, and 8B Instruct—available to the community at Groq speed. “This is so cool. It feels like AGI—you just talk to your computer and it does stuff instantly. Speed really makes AI so much more pleasing,” said former OpenAI researcher Andrej Karpathy, praising Groq's inference speed. Read the full story here. | | | | | | | | | | | | | Meanwhile, in India, Nidum AI announced a breakthrough in decentralised AI with its blockchain network enabling any device, ranging from smart fridges to Android phones and MacBook Pros to run Llama 3.1, allowing users to rent out compute power and earn rewards. AIM got in touch with the founder, Arjun Reddy, to understand the details of the project. Check it out here. While developers and startups are busy experimenting with Llama 3.1, Zuckerberg is using it to save his metaverse ambitions. “If you asked me five years ago, I would've guessed that AR would come before AI, not the other way around. So, some of this is just kind of setting yourselves up to ride the different waves when they come in,” he said, adding that metaverse was always a long-term thing. Check out the full story here. | | | | | | | | | | | Sleepless Nights for Google & Perplexity | | | | The truth of the matter: OpenAI's SearchGPT is finally here, and could dramatically transform internet search, directly challenging Google's $2 trillion dominance. By combining OpenAI’s advanced models such as GPT-4o with real-time web information, SearchGPT promises faster and more relevant answers to user queries. This new development sets the stage for a fierce competition in the search engine market. | | | | | | | | | | | | | | Google recently announced its upcoming hardware launch event, ‘Made by Google’, scheduled for August 13. The company has confirmed that it will unveil the Pixel 9, Pixel 9 Pro, and Pixel 9 Pro Fold at this event. Given the fierce competition, the tech giant is also likely to release the updated version of Gemma, aka Gemma 3, soon, which could be either at the event or before that. Meanwhile, Musk recently announced the launch of Grok 2 and 3, which according to him, is going to be the most powerful AI. It could potentially be on par or close to GPT-4. | | | | | | | | | | | Meanwhile, Mistral AI has released Mistral Large 2, which beats Llama 3.1 in coding and maths. With 123 billion parameters, this new model is designed for high efficiency and performance, setting new standards for open AI models. Its advanced function-calling capabilities and extensive language support make it a versatile tool for diverse applications. Check it out here. | | | | | Settling for Silver instead of Gold | | | | Recently, Google DeepMind’s AlphaProof and AlphaGeometry 2 AI models worked together to tackle questions from the International Math Olympiad (IMO). The DeepMind team scored 28 out of 42 – enough for a silver medal but one point short of gold. | | | | | | | | | | | | | | In response, one-third of the participants estimated that AI would achieve this level of success by 2025. The estimate isn’t way off to be honest. Bursting the bubble: Karpathy is dealing with a different kind of problem. He calls this “Jagged Intelligence”, which is described as the inconsistent performance of SOTA LLMs that excel at complex tasks, yet fail at simple ones, such as incorrectly determining that 9.11 is larger than 9.9. | | | | | | | | | | | In its second edition, MachineCon USA was grander and more impactful than last year. The event was successfully concluded on July 26 at the historic 583 Park Avenue in New York City, with over 350 AI and analytics leaders showcasing their innovations, sharing unique solutions, networking, and more. Additionally, AIM Research had the opportunity to honour thought leaders with the AI100 Award 2024, recognising and celebrating the achievements of individuals and organisations that have made significant advancements in the field of AI and analytics in enterprises. | | | | | | | | | | | Is OpenAI 'Sam' of All Trades, Master of None? OpenAI is treading on thin ice, with a recent report indicating potential losses of up to $5 billion for the company this year. OpenAI is expected to spend nearly $4 billion on renting Microsoft's servers to support ChatGPT and its underlying LLMs, including GPT-4o. The expenses for training and data acquisition could approach $3 billion, and with a rapidly growing workforce of around 1,500 employees, the operational costs could rise to approximately $1.5 billion. Overall, OpenAI's total operational expenses could reach up to $8.5 billion this year. At the same time, the company is struggling to find its moat. Read the full story here. 9-to-5 Jobs Will be Extinct by 2034 This isn’t about a 14-hour work day, or working for 70-hours a week. Reid Hoffman, the co-founder and executive chairman of LinkedIn, has made a startling prediction that the traditional 9-to-5 jobs will vanish by 2034. But unlike other bold claims that people keep making, this is not just another dystopian, or even utopian, proclamation. Read on. | | | | | | | | | | | | | | | | Masterclass on Optimising RAG Models for Enterprise-Grade Accuracy | | | AI Forum, in partnership with AIM & NVIDIA, is set to host a virtual workshop focusing on the optimisation of retrieval-augmented generation (RAG) models to achieve enterprise-grade accuracy. | | | Date: August 9, 2024 Time: 3:00 to 4:30 pm IST | | | | | | | | | | | | | - Researchers at NVIDIA have developed Llama3-ChatQA-2-70B, a large language model rivalling GPT-4-Turbo in handling contexts up to 128,000 tokens and excelling in RAG tasks.
| | | - Capgemini reported over 350 new projects and training 120,000 employees in generative AI, despite a 2.5% revenue decline in H1 2024.
| | | | | - Brev.dev, a San Francisco-based AI/ML development platform recently acquired by NVIDIA, launched Brev 2.5 to develop custom containers, enhancing flexibility and optimising GPU usage across various cloud providers.
| | | | | - GenAI startup Gnani AI, India's first voice-first SLM for vernacular languages, raised $3M from Info Edge to support its expansion plans.
| | | - Rabbitt AI secured a $2.1 million seed funding round led by TC Group, with participation from Meta, NVIDIA, and Microsoft executives, to advance its AI application development platform featuring custom LLMs, RAG fine-tuning, MLOps, and privacy-first AI deployment.
| | | - The Akash Ambani-backed AI startup, TWO, launched SUTRA Dual2L, a new multilingual large language model leveraging Llama 3.1 to support over 50 languages.
| | | | | | | | | | | | | Reach out to us at info@aimmediahouse.com Stay tuned for more insights in our next edition! | | We love hearing from our readers! Have thoughts on our new format? Questions, comments, or ideas are always welcome. If there’s a specific topic in AI or analytics that you're curious about, tell us! | | | | | | | | |
Комментариев нет:
Отправить комментарий
Примечание. Отправлять комментарии могут только участники этого блога.