Acuity – Acuity Inc. appoints Nick Tzitzon to its Board of Directors

EMR Analysis

More information on Acuity Inc.: See the full profile on EMR Executive Services

More information on Neil Ashe (Chairman, President and Chief Executive Officer, Acuity Inc.): See the full profile on EMR Executive Services

More information on Nick Tzitzon (Independent Director, Member of the Board of Directors and Member of the Audit and Governance Committees, Acuity Inc. + Vice Chairman, ServiceNow + Leader, AI Institute, ServiceNow): See the full profile on EMR Executive Services

More information on Karen J. Holcom (Senior Vice President and Chief Financial Officer, Acuity Inc.): See the full profile on EMR Executive Services

More information on ServiceNow: https://www.servicenow.com + SServiceNow (NYSE: NOW) is the AI control tower for business reinvention. The ServiceNow AI Platform integrates with any cloud, any model, and any data source to orchestrate how work flows across the enterprise. By unifying legacy systems, departmental tools, cloud applications, and AI agents, ServiceNow provides a single pane of glass that connects intelligence to execution across every corner of business. With more than 75 billion workflows running on the platform each year, ServiceNow helps organizations turn fragmented operations into coordinated, autonomous workflows that deliver measurable results. 

More information on Bill McDermott (Chairman and Chief Executive Officer, ServiceNow): https://www.servicenow.com/company/leadership.html + https://www.linkedin.com/in/billrmcdermott/ 

EMR Additional Notes:

• AI – Artificial Intelligence:
  • Artificial Intelligence (AI) is the broad field of computer science focused on developing systems that can perform tasks that typically require capabilities associated with human or other forms of intelligent behavior, such as learning, reasoning, perception, language understanding, planning, and decision-making.
  • AI systems may:
    • Process data or information from various sources;
    • Identify patterns, relationships, or relevant features;
    • Make predictions, classifications, decisions, or recommendations;
    • Reason, plan, generate content, or take actions toward defined goals.
  • AI is an umbrella term that includes machine learning, deep learning, generative AI, and other approaches such as rule-based systems, search, planning, knowledge representation, probabilistic methods, and optimization. Not all AI systems learn from data.
  • A popular but non-standard conceptual taxonomy describes AI as reactive machines, limited-memory systems, theory-of-mind AI, and self-aware AI. This taxonomy is useful for explaining different hypothetical levels of capability, but it is not a formal scientific classification of AI systems.
  • Main types of AI:
    • Type 1: Reactive machines. These AI systems do not use persistent memory of previous experiences to inform their decisions and generally respond only to the current input or state. An example is Deep Blue, the IBM chess-playing system that defeated Garry Kasparov in the 1990s. Deep Blue used chess-specific search and evaluation techniques to select moves rather than learning from previous games in the manner of modern machine-learning systems.
    • Type 2: Limited memory. These AI systems can use information from previous observations, stored data, or recent states when making decisions. Many modern AI systems use forms of memory or contextual information, although the term “limited memory” is a broad conceptual category rather than a precise technical classification.
    • Type 3: Theory of mind. A hypothetical form of AI capable of modeling aspects of other people’s beliefs, intentions, knowledge, emotions, or perspectives and using this understanding in interaction. Such capabilities remain an active area of research rather than an established category of deployed AI.
    • Type 4: Self-awareness. A hypothetical form of AI possessing a subjective sense of self or consciousness. No generally accepted evidence demonstrates the existence of genuinely self-aware or conscious AI systems.
  • AI programming therefore does not have a universally accepted set of “three cognitive skills”; learning, reasoning, perception, planning, language processing, and decision-making are among the capabilities commonly associated with AI.
  • Machine Learning (ML):
    • Machine Learning (ML) is a subset of AI in which algorithms learn patterns or relationships from data to perform tasks such as prediction, classification, generation, or decision-making, rather than relying solely on manually specified rules for every case.
    • ML uses historical or newly collected data to learn statistical patterns and improve performance on a defined task or objective.
    • ML is the dominant approach underlying many modern AI systems, but rule-based, optimization-based, search-based, and other non-ML approaches remain important in many applications.
    • ML allows software applications to learn from data and improve their performance on a defined task without requiring every decision rule to be explicitly programmed.
    • Recommendation engines are a common use case for ML. Other uses include fraud detection, spam filtering, image and speech recognition, forecasting, business process automation (BPA), and predictive maintenance.
    • Classical ML is often categorized according to how an algorithm learns from data. Common approaches include:
      • Supervised learning,
      • Unsupervised learning,
      • Semi-supervised learning,
      • Self-supervised learning, and
      • Reinforcement learning.
  • Deep Learning (DL):
    • Deep Learning (DL) is a subset of ML that uses neural networks with multiple computational layers to learn complex representations and relationships from data.
    • DL can be particularly effective for high-dimensional and unstructured or complex data such as images, audio, video, and natural language, although classical ML methods can outperform deep learning on some structured or tabular datasets.
    • DL uses multiple layers of information processing that can learn increasingly complex representations of input data. In some image-recognition systems, for example, earlier layers may learn lower-level visual features while later layers combine these features into more complex representations.
    • DL has enabled major advances in computer vision, speech recognition, natural language processing, generative AI, and other fields, but it is not inherently the “most sophisticated” AI architecture; different architectures and approaches are better suited to different tasks.
  • Generative AI (GenAI):
    • Generative AI (GenAI) refers to AI systems that generate new content or other outputs, such as text, images, code, audio, video, or structured data, based on patterns learned during training and on subsequent inputs or instructions.
    • GenAI is typically powered by machine-learning and deep-learning models, including large language models and other foundation models, rather than constituting a completely separate AI paradigm.
    • Generative AI can generate outputs based on prompts, structured inputs, multimodal inputs, tool calls, or instructions provided by people or software systems.
    • The public release of ChatGPT in November 2022 significantly increased public awareness and adoption of generative AI capable of producing natural-language responses and other forms of content.
  • Small Language Models (SLM) and Large Language Models (LLM):
    • Small Language Models (SLMs) are relatively compact language models designed to process and generate natural-language content, often with lower computational, memory, or deployment requirements than larger models.
    • Large Language Models (LLMs) are machine-learning models trained on large amounts of data to process and generate natural language and perform tasks such as text generation, summarization, translation, classification, question answering, reasoning, and information extraction.
    • “Small” and “large” are relative terms rather than universally defined categories based on a fixed number of parameters.
    • Modern LLMs are commonly based on transformer or related neural-network architectures.
    • Virtual assistants such as Siri, Alexa, and Google Assistant can incorporate language models, speech-recognition systems, search, rules, APIs, and other technologies; whether and to what extent they use LLMs depends on the specific version and implementation.
  • Computer Vision (CV) / Vision AI & Machine Vision (MV):
    • Computer Vision (CV) is a broad field of AI and computer science concerned with enabling computers to acquire, process, analyze, and interpret visual information from images, video, and other visual sensors.
    • CV can use deep learning, classical image-processing techniques, geometric methods, machine learning, and other computational approaches to recognize objects, detect patterns, estimate properties, understand scenes, or extract information from visual data.
    • Applications include object detection, image classification, optical character recognition (OCR), facial recognition, medical imaging, autonomous systems, visual search, and image-based measurement and inspection.
    • Machine Vision (MV) is an application of computer vision focused particularly on automated visual inspection, measurement, identification, guidance, and control, especially in industrial and manufacturing environments.
    • Machine Vision commonly uses cameras, lighting, image-processing software, AI or other vision algorithms, and interfaces to industrial control systems, robots, or other equipment.
    • The key difference is that CV is the broader technical field, while MV generally refers to application-oriented vision systems used to perform defined automated tasks.
  • Multimodal Intelligence and Agents:
    • Multimodal AI refers to AI systems that can process and/or generate information across multiple modalities, such as text, images, audio, video, and other sensor or structured data.
    • Multimodal capabilities allow AI systems to combine information from different types of input and output, enabling applications such as visual question answering, speech interaction, document understanding, and systems that can see, hear, speak, and generate visual or textual content.
    • An AI agent is a software system that perceives information from an environment, maintains or uses relevant state or context, and selects and takes actions toward a defined goal. An agent may incorporate one or more AI models, tools, memory, planning mechanisms, and orchestration components.
    • What distinguishes an AI agent from a conventional AI model is not simply decision-making, but the ability to take actions in an environment toward a goal, often through tools or external systems.
    • Agents can be classified or implemented in different ways, including reactive or planning-oriented agents, single-agent or multi-agent systems, and agents operating in static or dynamic environments. Multimodal agents can process multiple forms of information.
  • Agentic AI:
    • Agentic AI is an emerging term for AI systems designed to pursue defined goals by planning, making decisions, using tools, interacting with their environment, and taking actions with a degree of autonomy and limited human intervention.
    • Agentic AI systems may use large language models, other machine-learning models, planning mechanisms, memory or state, tools, external data sources, and orchestration components.
    • Unlike conventional AI models that primarily produce an output in response to an input, agentic systems extend model capabilities into action-oriented workflows, potentially executing multiple steps and adapting their actions based on intermediate results or changes in the environment.
    • Multi-agent systems are one implementation approach in which multiple specialized agents coordinate their activities to achieve a broader goal.
    • Agentic AI can build on generative AI, particularly LLMs, but agentic AI is not synonymous with generative AI or LLMs.
  • Agentic Enterprise:
    • An agentic enterprise is an emerging business operating model in which people, AI agents, enterprise applications, data, and automation systems work together to execute or coordinate business processes and workflows.
    • Compared with conventional automation, agentic systems can interpret information, make decisions within defined boundaries, use enterprise tools, and take actions across multiple workflow steps, while appropriate governance, security, permissions, monitoring, and human oversight can constrain or supervise those actions.
    • The term is increasingly used by technology companies and analysts to describe an evolution toward more autonomous AI-enabled enterprise operations, but it is not a standardized technical or organizational category.
  • Edge AI Technology:
    • Edge AI refers to the deployment and execution of AI models or inference workloads on or close to the devices and systems where data is generated, such as cameras, sensors, industrial controllers, vehicles, smartphones, and other IoT devices, rather than relying entirely on centralized cloud infrastructure.
    • Edge AI combines edge computing with AI/ML to process data locally or near its source, potentially reducing latency, bandwidth requirements, cloud dependence, or exposure of sensitive data.
    • Edge AI can operate with intermittent or no internet connectivity for some functions, although many systems continue to communicate with cloud or other centralized systems.
    • Applications include autonomous vehicles, industrial inspection, wearable devices, security cameras, robotics, and smart-home equipment.
  • High-Density AI:
    • High-density AI refers to the concentration of AI computing capacity—particularly GPUs or other accelerators, memory, networking, power delivery, and cooling—in a relatively small physical or data-center footprint.
    • It is an AI infrastructure and data-center design trend, not a distinct category of AI. High-density infrastructure can provide greater computing capacity per rack, room, or unit of floor space, but also creates significant requirements for power distribution, thermal management, networking, and physical infrastructure.
  • Explainable AI (XAI) and Human-Centered Explainable AI (HCXAI):
    • Explainable AI (XAI) refers to methods, techniques, and system designs intended to make AI model behavior, predictions, decisions, or outputs understandable to people.
    • Human-Centered Explainable AI (HCXAI) goes further by designing explanations around the needs, context, goals, knowledge, and capabilities of the people who use or are affected by the AI system.
    • While XAI focuses on understanding and communicating aspects of model behavior, HCXAI emphasizes whether explanations are relevant, comprehensible, useful, and appropriate for their human context. Related considerations can include fairness, accountability, trust, safety, and ethical use, although these are not themselves defining features of explainability.
  • Physical AI & Embodied AI:
    • Physical AI refers to AI systems designed to perceive, reason about, predict, and/or act in the physical world, often using data from sensors and controlling physical devices or actuators.
    • Embodied AI refers to AI systems whose intelligence is situated within or coupled to a physical or simulated body or agent, emphasizing the interaction between perception, decision-making, action, and the environment.
    • Physical AI and Embodied AI overlap substantially, and the terms are not universally defined as a strict hierarchy. They are commonly associated with robotics, autonomous vehicles, industrial systems, and other applications in which AI interacts with a physical environment.
  • Federated Learning and Reinforcement Learning:
    • Federated Learning (FL) is a machine-learning technique in which models are trained across multiple decentralized devices or organizations while the raw training data generally remains at its original location; model parameters, gradients, or other training information are communicated to an aggregation or coordination system. In simple terms: “Train AI without centrally collecting the raw data.”
    • Federated learning can improve data governance and reduce the need to centralize sensitive data, but it does not by itself guarantee privacy or security. Additional techniques such as secure aggregation or differential privacy may be used.
    • Reinforcement Learning (RL) is a type of machine learning in which an agent learns to select actions by interacting with an environment and receiving rewards or penalties, with the objective of improving its policy or long-term expected reward. In simple terms: “Learning through interaction and feedback.”
    • Federated Learning and Reinforcement Learning can be combined as Federated Reinforcement Learning (FRL), in which multiple agents or devices learn policies locally and share selected model information or updates to improve a common or coordinated policy without centrally sharing their raw experience data.
    • FRL is an active research area with applications being investigated in areas such as distributed resource management, communications networks, robotics, and autonomous systems.
  • AI Factories:
    • AI Factories are industrial-scale computing and operational infrastructures designed to transform data and computing resources into AI models, inference services, and other AI outputs at scale.
    • AI factories can encompass data preparation, model training, fine-tuning, inference, serving, storage, high-speed networking, accelerated computing, power, cooling, and associated software and operational systems.
    • “AI factory” is a conceptual and industry term rather than a standardized technical category. It can refer to a dedicated AI data center or, more broadly, an integrated infrastructure and operating model for producing and deploying AI capabilities at scale.
    • Companies and organizations use the term in different ways, including in connection with hyperscale and specialized data-center infrastructure, accelerated computing, and public-sector AI infrastructure initiatives.

• Hardware vs. Software vs. Firmware: 
  • Hardware is physical: it’s tangible electronic or mechanical components. It can break, wear out, or be damaged by environmental factors (heat, water, shock, etc.).
    • Since hardware is part of the “real” world, it all eventually wears out. Being a physical thing, it’s also possible to break it, drown it, overheat it, and otherwise expose it to the elements.
    • Here are some examples of hardware:
      • Smartphone
      • Tablet
      • Laptop
      • Desktop computer
      • Printer
      • Flash drive
      • Router
  • Software is virtual: it consists of programs and data that run on hardware to perform functions. It can be copied, modified, updated, or deleted.
    • Software is everything about your computer that isn’t hardware.
    • Here are some examples of software:
      • Operating systems like Windows 11 or iOS
      • Web browsers
      • Antivirus tools
      • Adobe Photoshop
      • Mobile apps
  • Firmware is virtual: is embedded software that is tightly coupled to specific hardware and controls its low-level functions.
    • While not as common a term as hardware or software, firmware is everywhere—on your smartphone, your PC’s motherboard, your camera, your headphones, and even your TV remote control.
    • Firmware is a specialized type of software that serves a specific control and interface role between hardware and higher-level software.