The world of GenAI/ LLMs/ LMMs are governed by 𝐓𝐨𝐤𝐞𝐧𝐬. Tokens are the building blocks of Language as a LLM perceives it- they can be whole words, parts of words, or even punctuation. The LLM economy transacts and charges per Token-In/Out. An LLM can be imagined, (in a reductionist way) as a Lego set. Where each Lego set comes in multiple different sizes of lego blocks, which when put together in a meaningful way creates appreciable outputs. Only but, the number of unique Lego blocks (tokens) can run up to few 100,000s, which together make meaningful generations. It is to be noted that the final collection of tokens that characterizes an LLM are unique to it and are optimized. Check out - (https://lnkd.in/dqgn5HUN) to play around with different tokenizers for each different GenAI models. The token sets per GenAI model is determined using 1) the algorithm that merges characters such as BytePairEncoding (used by GPT) or Sentencpiece (used by Google & Llama) and 2) Most critically/ importantly the 𝐭𝐫𝐚𝐢𝐧𝐢𝐧𝐠 𝐝𝐚𝐭𝐚𝐬𝐞𝐭𝐬. Most of the GenAIs models and their optimized tokens are derived from English Datasets (Mistral (french) and Jais (Arabic) are exceptions). Hence, when you try Tamil or Telugu on GPT-4 you will see that the GenAI consumes more tokens and making the usage extremely expensive in the token economy. GenAI models have set up themselves as indispensable tools towards societal growth. And it makes most sense for each cohesive societal unit that can adopt a Language Model to leverage it. It is in this is proactive thought and direction that the India AI mission is ushered in with a holistic view of establishing an Indian LLM stack - Compute Capacity, Indigenous Datasets & Models and Applications. Excited for the Brave New World Ahead ! |<EOS>| https://lnkd.in/ddYvUruf https://lnkd.in/dAwrhxvj
Sivas Subramaniyan’s Post
More Relevant Posts
-
Good article with great title! Indeed India has potential to be AI use-case capital of the world. In the world of fail fast system, sometimes it’s good to be a tortise who wins the race. Sustainable and Scalable implementation of AI specifically GenAI at reasonable cost is key to success more than any highly rated multi-billion parameter state of art model which may not justify the ROI.
India aims to be the AI use-case capital of the world, says Nandan Nilekani
moneycontrol.com
To view or add a comment, sign in
-
Great Insight
AI Leader | Successfully built AI culture within organisation | Building Impactful AI Enabled Products | CX-Ops | Fast Prototyping | Innovation Lab
Good article with great title! Indeed India has potential to be AI use-case capital of the world. In the world of fail fast system, sometimes it’s good to be a tortise who wins the race. Sustainable and Scalable implementation of AI specifically GenAI at reasonable cost is key to success more than any highly rated multi-billion parameter state of art model which may not justify the ROI.
India aims to be the AI use-case capital of the world, says Nandan Nilekani
moneycontrol.com
To view or add a comment, sign in
-
India is taking a big step towards becoming a leader in Artificial Intelligence with the approval of the India AI Mission. The government has allocated a budget of Rs.10,371.92 crore towards the mission, which aims to create a national AI ecosystem. The focus is on developing indigenous tools for the safe, trusted, and ethical deployment of AI. This mission is a significant stride towards unlocking AI's potential in India and establishing the country as a major player in this field. #IndiaAIMission #AIforIndia
Cabinet Approves Ambitious IndiaAI Mission to Strengthen the AI Innovation Ecosystem
pmindia.gov.in
To view or add a comment, sign in
-
Given the hype around generative AI is yet to translate into material sustainable revenue upside leading to transformed business models, the burden of cost had to give in. Next revelation, epiphany, what have you, is likely to be on the infra/compute side where organisations start questioning why do you need expensive GPUs for entirety of your AI usecase pipeline. There is more cost to be saved making experimentation rewarding and sustainable for both, the business and the planet. ET: Startups See their AI Bills Shrink as Global LLM Price War Escalates
Startups See their AI Bills Shrink as Global LLM Price War Escalates
economictimes.indiatimes.com
To view or add a comment, sign in
-
🚀 Exciting News in the AI World! 🚀 Today, OpenAI has released a new model for developers - GPT-4o Mini. This is a true game changer for small startups and beyond! The cost efficiency of this new model is astounding, being 3x cheaper than GPT-3.5 Turbo, more than 2x cheaper than Gemini-1.5 Flash, and 2x cheaper than Claude Haiku. Here are the main cost numbers: - GPT-4o Mini - $0.15 / 1M input tokens - $0.60 / 1M output tokens - Gemini Flash - $0.35 / 1M input tokens - $1.05 / 1M output tokens - Claude Haiku - $0.25 / 1M input tokens - $1.25 / 1M output tokens What does this mean for the industry? I believe Anthropic and Google will be compelled to decrease their prices for the next model generation, or at least match these prices. This competitive pricing is a win for everyone, leading to better services for end customers, startups, and enterprises alike. It's an exciting time to be in the AI space, and I'm eager to see how this will accelerate innovation and development. Read more here: https://lnkd.in/d8_DbUkw #AI #OpenAI #GPT4omini #Startups #Innovation #TechNews
GPT-4o mini: advancing cost-efficient intelligence
openai.com
To view or add a comment, sign in
-
About IndiaAI Mission: It will establish a comprehensive ecosystem catalyzing AI innovation through strategic programs and partnerships across the public and private sectors. It will be implemented by ‘IndiaAI’ Independent Business Division (IBD) under Digital India Corporation (DIC). It has the following components: IndiaAI Compute Capacity: It will build a high-end scalable AI computing ecosystem to cater to the increasing demands from India’s rapidly expanding AI start-ups and research ecosystem. It will comprise AI compute infrastructure of 10,000 or more Graphics Processing Units (GPUs), built through public-private partnership. It will act as a one-stop solution for resources critical for AI innovation. IndiaAI Innovation Centre: It will undertake the development and deployment of indigenous Large Multimodal Models (LMMs) and domain-specific foundational models in critical sectors. IndiaAI Datasets Platform: It will streamline access to quality non-personal datasets for AI Innovation. A unified data platform will be developed to provide a one-stop solution for seamless access to non-personal datasets to Indian Startups and Researchers. IndiaAI Application Development Initiative: It will promote the AI applications in critical sectors for the problem statements sourced from Central Ministries, State Departments, and other institutions. IndiaAI FutureSkills: It is conceptualized to mitigate barriers to entry into AI programs and will increase AI courses in undergraduate, masters-level, and Ph.D. programs. Further, Data and AI Labs will be set-up in Tier 2 and Tier 3 cities across India to impart foundational level courses. IndiaAI Startup Financing: It is conceptualized to support and accelerate deep-tech AI startups and provide them streamlined access to funding to enable futuristic AI Projects. Safe & Trusted AI: It will enable the implementation of Responsible AI projects including the development of indigenous tools and frameworks, self-assessment checklists for innovators and other guidelines and governance frameworks. The IndiaAI Mission will propel innovation and build domestic capacities to ensure the tech sovereignty of India.
Technophile II delivery-focussed II Avid Learner II IT Management II GAVP - DCB Bank (Head PMO & IT Governance , Head IT Procurement)
Recently, the Cabinet has approved the comprehensive national-level IndiaAI mission.
vajiramias.com
To view or add a comment, sign in
-
Indian genAI startup, Rabbitt.ai has announced the launch of ChanceRAG, a no-code Retrieval Augmented Generation (RAG) solution designed to simplify the integration of large language models (LLMs) with document retrieval systems. Harneet Singh, chief AI officer at Rabbitt.ai, highlighted the product as an “enterprise-grade solution for building RAG.” This new tool aims to help businesses and individuals easily integrate advanced AI-powered solutions into their workflows without needing any programming skills. RAG technology allows AI models to search and retrieve relevant data from external sources, making their responses more accurate and context-aware. With ChanceRAG, users can now leverage this technology without having to write code. The platform allows seamless connections to various data sources and databases, enabling users to generate precise answers or content based on up-to-date information. What sets ChanceRAG apart is its user-friendly interface, which simplifies the setup and management of AI models. Rabbitt.ai’s goal is to empower non-technical users to benefit from the latest AI advancements, improving productivity and decision-making processes across industries. With this new launch, Rabbitt.ai is making powerful AI tools more accessible, offering a streamlined solution to businesses looking to enhance their operations with cutting-edge technology. #ai #artificialintelligence #generativeai #nocodeai #largelanguagemodel #ainews #artificialintelligenceai #aitechnology #rabbittai #RAG #retrievalaugmentedgeneration #retrievalaugmented
To view or add a comment, sign in
-
Beijing-based Zhipu AI has raised 3 billion yuan ($411.8 million) in a funding round supported by state capital and strategic backers, including existing shareholders like Legend Capital. This investment aims to enhance the research and development of its language model, GLM, following a previous round valued at 20 billion yuan ($2.85 billion). As one of China’s "six tigers" in the large language model (LLM) sector, Zhipu AI has responded to the evolving generative AI landscape since the introduction of OpenAI's GPT-3. It launched the GLM-4 model in January, upgraded to GLM-4-0520, and released GLM-4-Plus in August. The company also introduced CogVideox, a text-to-video service, in July. In 2024, Zhipu AI reported 100% revenue growth, largely due to its Model as a Service (MaaS) platform, which has attracted over 700,000 businesses and developers. Its consumer app, Zhipu Qingyan, has 25 million users, generating annual recurring revenue exceeding 10 million yuan ($1.4 billion). Founded in 2019, Zhipu AI operates in the US, UK, France, Singapore, and Malaysia. It recently initiated a venture capital fund aiming to raise 1.5 billion yuan ($210.6 million) for early-stage LLM startups, supported by major venture capitalists and local tech giants like Tencent, Meituan, Xiaomi, and Alibaba. #AI #China Want more analysis of major developments in AI, semiconductors, and cloud computing in Asia? Subscribe now to the Cognitive Asia newsletter--free of charge--at Substack https://lnkd.in/g4s6bqfi https://lnkd.in/gziFxTMq
Chinese LLM unicorn Zhipu AI pockets over $411m
dealstreetasia.com
To view or add a comment, sign in
-
I agree with Nandan Nilekani: #India should harness its #frugal #innovation capabilities to make #AI not only cost-effective but also RELEVANT in Indian socio-cultural context characterized by great #diversity Check out Sarvam AI, a #Bengaluru-based #startup that just launched an open-source foundational model supporting 10 Indian languages, including #Hindi, #Tamil, and #Bengali: https://lnkd.in/gxupbfzs .
Frugal engineering can make AI cost-effective for India: Nilekani
thehindu.com
To view or add a comment, sign in
-
This is best reflected in projected CAGR growth of 37% for AI (until 2030). #dualuse #ai #artificialintelligence #techsolution #scaleups #growthcapital #development #returnoninvestment #investmentstrategies
Generative AI 'FOMO' is driving tech heavyweights to invest billions of dollars in startups
cnbc.com
To view or add a comment, sign in
Stay ahead of the curve with the evolving GenAI landscape! 🌐