AI Model Development

After a recent price reduction by OpenAI, GPT-4o tokens now cost $4 per million tokens (using a blended rate that assumes 80% input and 20% output tokens). GPT-4 cost $36 per million tokens at its initial release in March 2023. This price reduction over 17 months corresponds to about a 79% drop in price per year. As you can see, token prices are falling rapidly! One force that’s driving prices down is the release of open weights models such as Llama 3.1. If API providers, including startups Anyscale, Fireworks, Together AI, and some large cloud companies, do not have to worry about recouping the cost of developing a model, they can compete directly on price and a few other factors such as speed. Further, hardware innovations by companies such as Groq (a leading player in fast token generation), Samba Nova (which serves Llama 3.1 405B tokens at an impressive 114 tokens per second), and wafer-scale computation startup Cerebras (which just announced a new offering this week), as well as the semiconductor giants NVIDIA, AMD, Intel, and Qualcomm, will drive further price cuts. When building applications, I find it useful to design to where the technology is going rather than where it has been. Based on the technology roadmaps of multiple software and hardware companies — which include improved semiconductors, smaller models, and algorithmic innovation — I’m confident that token prices will continue to fall rapidly. This means that even if you build an agentic workload that isn’t entirely economical, falling token prices might make it economical at some point. Being able to process many tokens is particularly important for agentic workloads, which must call a model many times before generating a result. Further, even agentic workloads are already quite affordable for many applications. Let's say you build an application to assist a human worker, and it uses 100 tokens per second continuously: At $4/million tokens, you'd be spending only $1.44/hour – which is significantly lower than the minimum wage in the U.S. and many other countries. So how can AI companies prepare? - First, I continue to hear from teams that are surprised to find out how cheap LLM usage is when they actually work through cost calculations. For many applications, it isn’t worth too much effort to optimize the cost. So first and foremost, I advise teams to focus on building a useful application rather than on optimizing LLM costs. - Second, even if an application is marginally too expensive to run today, it may be worth deploying in anticipation of lower prices. - Finally, as new models get released, it might be worthwhile to periodically examine an application to decide whether to switch to a new model either from the same provider (such as switching from GPT-4 to GPT-4o-2024-08-06) or a different provider, to take advantage of falling prices and/or increased capabilities. [Reached lenght limit. Full text: https://lnkd.in/gz-xffF4 ]

𝗠𝗼𝘀𝘁 𝗰𝗼𝗺𝗽𝗮𝗻𝗶𝗲𝘀 𝗯𝗲𝗹𝗶𝗲𝘃𝗲 𝘁𝗵𝗮𝘁 𝗔𝗜 𝗶𝘀 𝗮 𝘀𝘁𝗿𝗮𝗶𝗴𝗵𝘁 𝗽𝗮𝘁𝗵 𝗳𝗿𝗼𝗺 𝗱𝗮𝘁𝗮 𝘁𝗼 𝘃𝗮𝗹𝘂𝗲. The assumption: 𝗗𝗮𝘁𝗮 → 𝗔I → 𝗩𝗮𝗹𝘂𝗲 But in real-world enterprise settings, the process is significantly more complex, requiring multiple layers of engineering, science, and governance. Here’s what it actually takes: 𝗗𝗮𝘁𝗮 • Begins with selection, sourcing, and synthesis. The quality, consistency, and context of the data directly impact the model’s performance. 𝗗𝗮𝘁𝗮 𝗦𝗰𝗶𝗲𝗻𝗰𝗲 • 𝗗𝗮𝘁𝗮 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴: Exploration, cleaning, normalization, and feature engineering are critical before modeling begins. These steps form the foundation of every AI workflow. • 𝗠𝗼𝗱𝗲𝗹𝗶𝗻𝗴: This includes model selection, training, evaluation, and tuning. Without rigorous evaluation, even the best algorithms will fail to generalize. 𝗢𝗽𝗲𝗿𝗮𝘁𝗶𝗼𝗻𝗮𝗹𝗶𝘇𝗮𝘁𝗶𝗼𝗻 • Getting models into production requires deployment, monitoring, and retraining. This is where many teams struggle—moving from prototype to production-grade systems that scale. 𝗖𝗼𝗻𝘀𝘁𝗿𝗮𝗶𝗻𝘁𝘀 • Legal regulations, ethical transparency, historical bias, and security concerns aren’t optional. They shape architecture, workflows, and responsibilities from the ground up. 𝗔𝗜 𝗶𝘀 𝗻𝗼𝘁 𝗺𝗮𝗴𝗶𝗰. 𝗜𝘁’𝘀 𝗮𝗻 𝗲𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 𝗱𝗶𝘀𝗰𝗶𝗽𝗹𝗶𝗻𝗲 𝘄𝗶𝘁𝗵 𝘀𝗰𝗶𝗲𝗻𝘁𝗶𝗳𝗶𝗰 𝗿𝗶𝗴𝗼𝗿 𝗮𝗻𝗱 𝗼𝗽𝗲𝗿𝗮𝘁𝗶𝗼𝗻𝗮𝗹 𝗺𝗮𝘁𝘂𝗿𝗶𝘁𝘆. Understanding this distinction is the first step toward building AI systems that are responsible, sustainable, and capable of delivering long-term value.

Every board is betting big on AI. Almost none are asking the question that actually protects them. I’ve been in boardrooms across industries, from finance to healthcare, and I keep seeing the same thing: Board members ask: - “What’s the AI budget?” - “What’s the timeline?” - “What’s the ROI?” But almost no one asks the most important question: “How do we even know this is AI?” Here’s the problem… Most boards are approving AI initiatives without a clear definition of what qualifies as AI because the lines are blurry. Vendors show up with polished demos and pitch tools labeled “AI-powered.” But without clarity, boards end up greenlighting: ✗ Rule-based systems dressed up as intelligence ✗ Traditional software relabeled with buzzwords ✗ Proof-of-concept demos, not scalable AI infrastructure ✗ “AI-washed” features that don’t actually learn or adapt Before the next AI contract crosses your desk, ask leadership: → Where exactly does machine learning happen in this system? → How does it improve over time with use? → What data powers it, and who owns that data? → How much human intervention is required for results? Because the companies truly win with AI? They’re not the ones with the flashiest tools. They’re the ones whose boards can differentiate real intelligence from noise. What’s your take - have you seen “AI” claims fall apart under scrutiny?

The Death of SaaS (as We Know It) Satya Nadella recently shared a fascinating perspective: AI is poised to replace traditional application layers, embedding business logic directly at the database level. This marks a profound shift: one that could redefine the very foundation of SaaS. Imagine a future where AI doesn’t just power apps but replaces them. Business logic, instead of flowing through multiple layers of UI, middleware, and APIs, is orchestrated directly with the database. This means the end of bloated, layered software and the beginning of lean, AI-native architectures. The ripple effects are massive. SaaS as a subscription model may lose relevance as modular AI-driven workflows dominate. Interfaces will transform, shifting away from dashboards and fixed workflows to adaptive, real-time experiences—think voice commands, conversational AI, or neural interfaces. Even the app store economy may collapse under the weight of this new paradigm, replaced by marketplaces for AI-driven workflows instead of apps. This could imply the extinction for the SaaS we know today. For developers, businesses, and consumers, this shift will reshape how software is built, sold, and used. The question isn’t if SaaS is dying; it’s what comes next. What do you think? Is this the end of SaaS, or the beginning of something even more disruptive?

They tell you training and running AI model costs billions. That's true for a few frontier labs. But for most real-world use cases? Dramatically lower than you think thanks to open-source. Real examples from @HuggingFace's latest analysis: - Fine-tune a text classification model: <$2k - Train a leading image embedding model: < $7k - Train Deepseek OCR: < $100k - Train a leading machine translation model: <$500k Compare that to GPT-4.5 training (~$300M est.) And the truth is that you don't need a Formula 1 car to pick up groceries. Most tasks are solved just as well by smaller, efficient, targeted models. The mistake everyone makes? Starting with "what's the best AI model?" instead of "what do I need to do?" The future of AI is not just bigger models. It's cheaper, more customized, open models solving specific problems. Explore 100+ real model costs of training and deployment yourself in the study!

AI-assisted coding isn’t just about autocomplete anymore. It’s becoming a full lifecycle - from planning to building to reviewing. Developers are no longer just writing code, they’re orchestrating systems of agents that generate, test, and refine it. The shift is from “write code faster” to “build and ship systems end-to-end.” Here’s how the generative programmer stack is evolving 👇 𝗕𝗨𝗜𝗟𝗗 - 𝗖𝗼𝗱𝗲 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝗼𝗻 & 𝗘𝘅𝗲𝗰𝘂𝘁𝗶𝗼𝗻 Full-Stack App Builders: Turn ideas into working applications quickly by generating frontend, backend, and integrations in one flow. CLI-Native Agents: Work directly from the terminal to generate, edit, and execute code with tight control and speed. IDE-Native Agents: Integrate inside development environments to assist with coding, debugging, and real-time suggestions. Async Cloud Coding Agents: Run tasks in the background - writing, testing, and iterating on code without blocking your workflow. 𝗣𝗟𝗔𝗡 - 𝗣𝗹𝗮𝗻𝗻𝗶𝗻𝗴 & 𝗙𝗲𝗮𝘁𝘂𝗿𝗲 𝗕𝘂𝗶𝗹𝗱𝗶𝗻𝗴 Spec-first Tools: Start with structured specifications that define what to build before writing any code. Ask / Plan Modes: Break down problems, explore approaches, and validate logic before jumping into implementation. Design-to-Code Inputs: Convert designs or structured inputs into working code, reducing manual translation effort. 𝗥𝗘𝗩𝗜𝗘𝗪 - 𝗥𝗲𝘃𝗶𝗲𝘄, 𝗧𝗲𝘀𝘁𝗶𝗻𝗴 & 𝗩𝗲𝗿𝗶𝗳𝗶𝗰𝗮𝘁𝗶𝗼𝗻 Code Review Agents: Automatically analyze code for issues, improvements, and best practices before deployment. Testing & Verification: Generate and run tests to ensure reliability, correctness, and stability across different scenarios. Benchmarks: Measure performance and quality using standardized evaluation frameworks. What this means: Coding is shifting from manual effort to guided execution. The developer’s role is moving toward direction, validation, and system design. The edge is no longer just writing better code. It’s knowing how to use these tools together to ship faster and more reliably. Which part of this workflow are you using AI for the most today?

For the past year, I’ve had the privilege of co-chairing together with Carme Artigas the UN’s High-level Advisory Body on AI,which included 38 members from 33 countries. We were tasked with developing a blueprint for sharing AI’s transformative potential globally, while identifying and addressing the risks and filling the gaps that limit participation.  Following our interim report in Dec 2023, today we’re sharing our final report which outlines our key findings and recommendations to enhance global cooperation on AI governance. The report was informed by extensive consultation, including more than 2000 participants from all regions, 18 deep dives with 500 expert participants, 250 written submissions, 100+ virtual discussions, as well as research and surveys. AI has the potential to assist people in everyday tasks to their productive and creative endeavors, enable entrepreneurs and small and large businesses, transformation of sectors from healthcare to agriculture, power economic growth, advance science in ways that benefit society, and contribute to achieving the UN’s Sustainable Development Goals. At the same time, as with any powerful technology, it poses risks, challenges and complexities ranging from bias, misapplication and misuse, impact on work, to potentially widening global inequities. Our work highlighted many of these themes as well as key gaps in governance and the capacity for all to fully benefit from AI.  To harness AI’s potential and mitigate its risks, we need a truly inclusive and international effort – and current governance structures are missing too many voices. Our recommendations focus on these and other findings and I encourage you to read the report. Thank you to the UN’s Tech Envoy Amandeep Gill and his team, my co-chair Carme Artigas, and my fellow members of the advisory body -- from whom I learned a lot -- for their expertise and diverse views and vantage points, partnership, persistence and commitment to governing and harnessing AI’s potential benefits for all of humanity. https://lnkd.in/gFhFWWEh Carme Artigas, Anna Christmann, Anna Abramova, Omar Sultan AlOlama, @Latifa Al-Abdulkarim, Estela Aranha, Ran Balicer, Paolo Benanti, Abeba Birhane, Ian Bremmer, Anna Christmann,Natasha Crampton, Nighat Dad, Vilas Dhar, Virginia Dignum, @Arisa Ema, @mohamed farahat, Wendy Hall, Rahaf Harfoush, Hiroaki Kitano, Haksoo Ko, Andreas Krause, Maria Vanina Martinez, Seydina M. Ndiaye, @Moussa Ndiaye, Mira Murati, Petri Myllymäki, Alondra Nelson, Nazneen Rajani, Craig Ramlal, @Ruimin He, Emma Ruttkamp-Bloem, Marietje Schaake, @Sharad Sharma, @Jaan Tallinn, Ambassador Philip Thigo, MBS, Jimena Viveros LL.M., Yi Zeng, @Zhang Linghan

AI Engineering has four levels to it! – Level 1: Using AI Start by mastering the fundamentals: -- Prompt engineering (zero-shot, few-shot, chain-of-thought) -- Calling APIs (OpenAI, Anthropic, Cohere, Hugging Face) -- Understanding tokens, context windows, and parameters (temperature, top-p) With just these basics, you can already solve real problems. – Level 2: Integrating AI Move from using AI to building with it: -- Retrieval Augmented Generation (RAG) with vector databases (Pinecone, FAISS, Weaviate, Milvus) -- Embeddings and similarity search (cosine, Euclidean, dot product) -- Caching and batching for cost and latency improvements -- Agents and tool use (safe function calling, API orchestration) This is the foundation of most modern AI products. – Level 3: Engineering AI Systems Level up from prototypes to production-ready systems: -- Fine-tuning vs instruction-tuning vs RLHF (know when each applies) -- Guardrails for safety and compliance (filters, validators, adversarial testing) -- Multi-model architectures (LLMs + smaller specialized models) -- Evaluation frameworks (BLEU, ROUGE, perplexity, win-rates, human evals) Here’s where you shift from “it works” to “it works reliably.” – Level 4: Optimizing AI at Scale Finally, learn how to run AI systems efficiently and responsibly: -- Distributed inference (vLLM, Ray Serve, Hugging Face TGI) -- Managing context length and memory (chunking, summarization, attention strategies) -- Balancing cost vs performance (open-source vs proprietary tradeoffs) -- Privacy, compliance, and governance (PII redaction, SOC2, HIPAA, GDPR) At this stage, you’re not just building AI—you’re designing systems that scale in the real world. What else would you add? Subscribe to my free blog for more learning blog.dataexpert.io

"How do we ensure that the rapid development of AI is more considerate of harms and the public interest? In our inaugural Responsible AI Impact Report, All Tech Is Human (ATIH) aims to reveal our most urgent risks, emerging safeguards, and public-interest solutions, and provide a roadmap for how we will shape how AI impacts society in the year ahead. We examine the state of Responsible AI (RAI) throughout 2025 and highlight what we consider to be some of the most impactful contributions made by civil society organizations this year to enrich this broad and dynamic field. We believe the Responsible AI field can only thrive if we effectively tackle the complex challenges at the intersection of technology and society. When we refer to “Responsible AI,” we mean AI that is well-regulated and guard-railed, governed and assured (documented, standardized, and benchmarked with relevant measurements), and assessed, evaluated, and red-teamed. As we outlined in our recent Responsible Tech Guide (2025), our organization believes in a human-centered future that values our agency in desired outcomes and rejects tech determinism. As such, we are focused on elevating AI models that do as little harm as possible, for use cases in which risks have been carefully considered and meaningfully mitigated; and ethically deployed AI, in which lofty principles are operationalized with grounded KPIs. This Responsible AI Impact Report highlights the growing focus on Public Interest AI that is of, by, for, and in service to the people. This Public Interest AI should be applied to humanity’s most pressing challenges and enable us to reimagine what a better tech future entails. This report also explores a future in which Public Interest AI is developed on public infrastructures for an AI-literate society. At the heart of the years ahead lies a defining question: who determines the purpose of AI and the kinds of lives it will shape?" Rebekah Tweed, with support from David Ryan Polgar, Sandra Khalil, and Sherine Kazim. 

"Inclusive AI" - a noble goal or a new front of AI geopolitics?   When we say #inclusiveAI, we generally mean AI systems designed to be non-discriminatory, unbiased and accessible (particularly for marginalised and underrepresented groups), via data cleaning techniques, diverse decision making process, human oversight, etc.   But what does "inclusive AI" mean on a global scale?   Typical answers include: ✅ Equitable access to AI and their benefits across different countries and regions. ✅ Incorporating diverse perspectives and cultural contexts into AI development and deployment. ✅ Using AI to address global challenges (e.g. climate change).   But recently, I've been thinking a lot about inclusive AI, especially as "inclusiveness" becomes increasingly mentioned in international forums. For example (just from November 2024 alone): ▪ the Joint Ministerial Statement of the recent APEC summit (at Peru) mentions "inclusive" 24 times ▪ the G20 Rio de Janeiro Leaders' Declaration (at Brazil) also mentions "inclusive" 23 times.   At this month's G20 summit, China President Xi delivered a speech, calling for the G20 to ensure AI is "for good and for all, not a game of the rich countries and the wealthy" (see news headlines).   I'd count this as another reference to inclusive AI, particularly around rebalancing access to AI between the "rich" Global North and the Global South.   That brings us to the #geopolitics of things.   The reality is that AI is only the tip of the iceberg. It's intertwined with other 'deeper' areas like cloud, telco infrastructure, cross-border data and funds transfer, chips, international payments, big tech presence, start-up environment etc. There would need to be "inclusiveness" in those areas for us to achieve some real level of globally inclusive AI. Yet each area has its own practical and geopolitical challenges. Also, it depends on whether you're looking at "inclusive AI" from the perspective of consumer access, development & innovation capabilities, smart infrastructure, etc. Each angle tells a different story.   Given the realities, do you ever wonder if "inclusive AI": ❓ is now just another codename for "sovereign AI" or "AI nationalism" (i.e. each nation adopts policies/laws that advance their own domestic AI capabilities while also hampering overseas competitors)? It's arguably "inclusive" because each nation promotes AI access for their own population... ❓will create a 'tokenistic' mindset to AI where efforts towards inclusive AI are superficial, focusing on surface-level diversity rather than addressing deeper systemic issues? ❓will promote a culture of data exploitation as more data is collected under the guise of diversity/inclusion? Related concepts include "technofeudalism", "AI/digital colonialism", etc.   What do you think? 👓 Want more? I track #AI laws and policies around the world in my Global AI Regulation Tracker (see link in the 'Visit my website' button above). #tech #aiethics