ShiSh S.’s Post

For AI Giants, Smaller Is Sometimes Better EXCERPT: The start of the artificial-intelligence arms race was all about going big: Giant models trained on mountains of data, attempting to mimic human-level intelligence. Now, tech giants and startups are thinking smaller as they slim down AI software to make it cheaper, faster and more specialized. This category of AI software—called small or medium language models—is trained on less data and often designed for specific tasks. The largest models, like OpenAI’s GPT-4, cost more than $100 million to develop and use more than one trillion parameters, a measurement of their size. Smaller models are often trained on narrower data sets—just on legal issues, for example—and can cost less than $10 million to train, using fewer than 10 billion parameters. The smaller models also use less computing power, and thus cost less, to respond to each query.... #techgiantsthinksmaller #smallercheaperaimodels #aimodelstentimescheaper

Simple AI - A Good Bet Tom Dotan and Deepa Seetharaman provide solid coverage of the trends unfolding in AI Model development >> meets commercial reality. Key Points: > AI giants like Microsoft and Google are betting 10s of billions 💰 on #GenAi; now they are waking up to the enormous operating costs of training and running Large Language Models (#llms) > Top executives have announced scaling back opex burn associated with the new technology, as they start to realize that there is a disconnect between the vision and current commercial realities > Enterprise and Government customers are interested in AI as a way to get more value out of #data and #bigdata > Yet, practitioners and business operators are less interested in expensive 'wow' factor and want to understand ROI > GenAI makes a mockery of many ESG agendas, which have been publicly promoted by BigTech; energy hungry datacenters are proliferating and executives wonder if we can run GenAI, charge EVs, and keep the lights on at the same time Suggestions: 🎲 If you are betting on AI, be smart and follow the advice of Einstein who said, "make things as simple as possible, but not more so." 🎆 We're still a long way off of General Artificial Intelligence. Robots with AI brains are not about to walk into your workplace and replace your team. Instead, look for practical AI-driven solutions that multiply the value of your team! 🤖 AI is already advanced enough to make a huge impact in the enterprise. To win, keep it simple and aim the technology at well known problems. Partner with young innovative companies who are native AI thinkers and innovators. Hit us up at Terzo if you are considering Enterprise AI for Finance and Financial Intelligence. #cfo #ceo #finance #analytics Andy Zschach Spencer Ross Jake Flaherty brody elkins

⛳ Time for change in AI approach - 'surgical strike' v/s ‘land grab’ ? We often face a choice - Choice to accept a service that has colossal appeal with omni-presence, yet does a mediocre (passing marks) job versus choosing a targeted solution, available in pockets that solves specific problems, but it does it really well. Global corporations investing in AI are conflicted with similar choices. AI is facing a moment of truth to go big and wide or inculcate a sharp edge. Under the hood, achieving maturity and training for large models is super expensive. for example OpenAI GPT4.0 will take USD 100 mn to build and minimum of a trillion data records. Instead small and niche models are paving way to real problem-solving on ground. AI companies ranging from Microsoft to Mistral AI, Anthropic and Cohere are all focusing on surgical strike in AI. For example a legal issue AI may take only 10 million and a billion date points with significantly lower costs and for thsoe reasons Microsoft is building model at 1/100 th of scale of GPT4.0 Similarly DigiAlly's sharp edged AI/ML models are focused on addressing access to finance with target segments such as banks, financial institutions and supply chain for large corporations. We address the key question of trustworthiness, credibility of the client, mapped to industry benchmark with predictive capabilities giving them higher accuracy and efficiency. On the other hand, generic credit score (that goes stale the moment it is created), a prevalent practice in credit world, hasn't evolved at all. #ai #artificialintelligence #machinelearning #google #genai #digially Google #wsj Ajit Jaokar Vikram Pandya Prasanna Lohar Nitin Agrawal

LET'S GET SMALL (with #AI): It's obvious in my mind, but I am always focused on operationalization and meeting model / AI #governance requirements; small models are much more congruent with that, particularly since one of my Zoldi-isms is "explainability > predictive power." Perhaps other organizations will back into that belief as a byproduct of adopting smaller AI models; then, perhaps the realization will be that when we no longer need the AI giants, small language models will be within the grasp of internal AI teams. Edited to add: This was one of my #AIpredicitons2024, in fact. 2. Small will be beautiful "Large AI models, including large language models (LLMs) like ChatGPT, Google Bard AI, et. al., are incredibly unwieldy, difficult and expensive to build and operate. Furthermore, they are not easily retrained without cannibalistic entropy. "I have long held an unorthodox belief in “explainability first, predictive power second,” a core tenet of Responsible AI. In 2024, small, highly explainable AI models will be accepted and embraced as being more effective than outsized large models, as organizations focus on practical, reliable, net-positive business outcomes." My full predictions blog is here: https://lnkd.in/gbuVMVfN

AI in better size

A good synopsis. Lots of value can be delivered through smaller models focused on specific problems and domains.

Discover the Future of AI: OctoAI's Game-Changing Move in the LLM Arena! 🚀 🔥Breaking News in AI: Mixtral's collaboration with OctoAI is setting new benchmarks! Why is the tech world buzzing about it? Let's dive in. 💡 The Challenge: OpenAI has been a pioneer, but as LLM-powered apps skyrocket in popularity, innovators are seeking more. Why? The road to a successful AI-driven business is paved with high inference costs and limitations of closed-sourced models. 🚀 Enter OctoAI: Offering the fastest, most cost-effective solution to run Mixtral 8x7B. It's not just about speed; it's about efficiency and accessibility. 💼 Business Impact: Imagine slashing your costs by 12 times and boosting speed by 30% compared to GPT-4-Turbo. That's the reality with OctoAI. For startups and established businesses alike, this means unparalleled growth opportunities without the hefty investment. 👩💻 Seamless Transition: The best part? Switching to OctoAI is a breeze. Their API format mirrors OpenAI's, ensuring minimal engineering effort. It's about making your life easier, not harder. 📈 Proven Results: Customers making the switch are witnessing comparable quality with significant savings and lower latency. It's a win-win. 🎁 Exclusive Offer: Ready to experience the revolution? Try OctoAI Text Generation and get a $10 credit on signup. It's time to transform your AI strategy! Embrace the new era of AI with OctoAI and Mixtral. Share your thoughts below and let's discuss the future of AI! #OctoAI #Mixtral #AIInnovation #TechTrends #BusinessGrowth

Yay! 🎉 Snorkel AI got mention in recent The Wall Street Journal article. Our unique programmatic approach is helping companies customize their AI models 10-100 times faster by embedding their business expertise. As we explore the practical applications of large language models (LLMs), we’re turning our attention to specialized models and smaller models. These not only reduce latency but also provide immense value. To understand the state of GenAI… Read more here 👇

“Generative AI does 3 things. 👍 Interface: GenAI interfaces detect intent and break it down into tasks. 👍 Access: GenAI can provide access to data, models, or other apps based on an intent or well-defined outcome. 👍 Orchestration: GenAI can orchestrate interactions between people and multiple agents, apps, and models to facilitate a complex workflow.” Vin Vashishta great summary on #genai capabilities

GROQ: The Game-Changing AI Accelerator? Picture this: you're working on a cutting-edge AI project, and you need lightning-fast processing speed. Enter GROQ, the AI accelerator that's about to revolutionize the game. As a technology enthusiast and advocate, I can't wait to get my hands on the commercial version of this incredible innovation. The Need for Speed: What Makes GROQ Special? So, what exactly is GROQ, and why is it causing such a buzz? Imagine an AI accelerator designed to tackle the most demanding machine learning and AI workloads head-on. GROQ's architecture is built for maximum throughput and minimum latency, making it a dream come true for industries that rely on real-time data processing. The Balance of Speed and Quality Now, you might be thinking, "Speed is great, but what about quality?" It's a valid concern. Take Perplexity, for example. It's a modern Large Language Model that boasts impressive speed, but in my experience, it sometimes falls short in terms of output quality compared to GPT-4. The key is finding the right balance between speed and accuracy. Cost and Scalability: The Road Ahead I'll be honest; GROQ hardware isn't cheap. But as production scales up, we can expect prices to become more competitive. It's a familiar story in the world of advanced technology, and I have no doubt that GROQ will follow suit. Overcoming Context Length Limitations One potential drawback of GROQ is its limitation on the length of context it can handle. But fear not! Software solutions are available to manage this issue, although their effectiveness may vary depending on the task at hand. The Ultimate Showdown: Speed vs. Flexibility At the end of the day, the choice between speed and flexibility is a nuanced one. While GROQ offers mind-blowing processing speed, its limitations and costs can't be ignored. But let's not forget the true king of AI success: high-quality training data. As we eagerly await the commercial release of GROQ, one thing is clear: the future of AI is bright, and GROQ is poised to play a significant role in shaping it. Stay tuned as I continue to explore and share insights on the latest advancements in this exciting field! #MachineLearning #AIAccelerator #TechInnovation #FutureOfAI