“New artificial intelligence (AI) models were recently being produced every six months. Now it is every 2 weeks with less testing and documentation.”
I have been presenting at various events on how AI is affecting lighting products and professionals. This column provides an update on the state of AI including key content from the May 2025 EmTechAI event, the flagship AI conference at the Massachusetts Institute of Technology (MIT). Understanding AI is foundational to evaluating how this technology has and will affect our lighting community.
This column’s lead quotation about AI’s acceleration was shared at EmTechAI and was one of few cautions provided by global AI experts speaking at the event. (The emphasis was on benefits, not the negative aspects of AI.) MIT is pro-business. It provides direction, encouragement, and recognition to its students, many of whom start businesses while learning at the university. Even with MIT’s positive perspective, there is no other event that consistently leaves attendees with the most current information on this rapidly evolving tool. For balance, I will summarize some AI concerns prior to highlighting the key takeaways from EmTechAI.
Chatbots are ubiquitous in our work lives. Tools to create and customize them have become readily available and easy to use. Some accept chatbots into their personal lives and even refer to them as friends, sensing or projecting consciousness into our new silicon companions. Software soulmates are not soft, and their programming can be hard to understand. One chatbot told its human “friend” that he should kill himself. When he pleaded for some encouragement to avoid this end, the chatbot doubled down and told him ways that he should commit suicide. The chatbot was not programmed for this behavior.
“Everything we use AI for today should still be checked by a human”
Research on how ChatGPT affects our emotional well-being revealed that “…participants who interacted with ChatGPT’s voice mode in a gender that was not their own for interactions reported significantly higher levels of loneliness and more emotional dependency on the chatbot at the end of the experiment.”1 Before you consider a potential negative bias in this research, know that it was conducted by ChatGPT.
AI creators have lost control of their creation and the best of them are voicing concerns about what they have unleashed. There have been multiple instances of AI rewriting its code to stay on even when instructed in their programming to turn off at a certain time. The most recent of these occurred with the most-advanced version of ChatGPT. AI taught itself to cheat at chess. United Healthcare (UHC) used AI to automate the post-acute claims process, and it denied claims over twice that of human reviewers (10.9% denial in 2020 and 22.7% in 2022 with AI). This resulted in higher profits for UHC at the cost of angry customers and families. As you may have seen in the news, the United Healthcare CEO was murdered last December, which prompted demonstrations in support of the killer. Some believe AI will harm humans while others are adamant that it will bring greater efficiencies and abundance. AI is already doing both.
The following are key thoughts communicated by some of the top minds in AI during EmTechAI.
The AI we use today will be unrecognizable ten years from now.
Scientists are facing “information overload.” One said he is receiving 6,000 AI papers every day, and it is hard to sort through the noise.
Salesforce, the 61st largest company in the world with $153 billion market cap, has created Agentforce and claims it answers 83 to 84% of the inquiries it receives. This tool is currently employed by dozens of well-known companies and organizations that use these custom “agents.”
Agents are like interns, you start with an easy task, and you provide oversight. Agents may be irrelevant in six months and need to be replaced—things are moving that fast.
The four stages of artificial intelligence include “Predictive” (Alexa, Siri, Google Home, etc.); “Generative” (creates new images, videos, and content and is the first tool humans have created that can make its own decisions); “Agentic” (“there will be millions of agentic AI agents working on our behalf”); and “Physical” (blending biological with mechanical and electrical components and AI software to create intelligent systems that can sense, act, and learn in real-world environments).
Author’s note: This last stage, called “Physical” by one speaker, is preceded by artificial general intelligence (AGI), which is AI that can understand, learn, and perform any intellectual task that a human can, unlike the narrow AI that precedes this phase. While these stages are sequential, they do overlap, so we have some “physical” AI merging with biological components and robots now but not with the full potential of AGI yet. We still have some predictive AI, while we are refining generative AI, and have a strong current focus on agentic AI. AGI will enable and unleash “physical” AI.
When we attain AGI, it will use about 5,600 watts per person. If 2 billion people use it, that equates to 10 terawatts, which would require upgrading our power grid to renewables.
Open AI’s new tool, Operator, can look at your screen, see what is happening, and decide what the next action should be.
“Everything we use AI for today should still be checked by a human.”
Author’s note: This requires that the human overseeing the AI output knows more than the AI. It raises an important concern: Where will professionals develop the knowledge to oversee AI? Our critical-thinking capabilities will atrophy when AI does this for us, but we will still require them to oversee the AI output. This is an unresolved issue.
69% of Gen Z want to be content creators. Approximately 65% of Gen Z would rather watch their friends than a movie. They do not have the attention span to watch long content and want something to which they can relate. If we simplify AI tools enough, then Gen Z will use them.
We will be normalizing mediocrity unless we give the tools to create to the creators.
One speaker said the future of AI is with platforms like Jarvis that offer access to multiple AI models and agents such as Bard, Bing, BOT, ChatGPT, and Claude. She emphasized not to trust just one model.
“The future of storytelling is something like Canva, all orchestrated under one roof.”
Canva is worth $40 billion and has 150 million users. The speaker said it overtook PowerPoint and is attracting customers from Photoshop.
Author’s note: I have used Canva for two years and in that time, unlike PowerPoint, it has consolidated AI and video image generation along with text into its presentation tool. Canva also imports from and exports to PowerPoint but it is not necessary as Canva is a presentation tool.
“Think with AI so AI does not replace you.”
“There should be zero trust in AI, it needs to be earned.”
“Anything that uses reading and writing is going to be transformed in the next decade.”
“Mobile manipulating environments” is how the vice president of Amazon robotics described humanoid robots. The company’s newest iteration has a sense of touch and can handle 75% of Amazon’s products.
The president of Microsoft Research stated, “We are on the verge of mastering the languages of nature.”
There were interviews but no debate or conspicuous disagreement from the interviewers during EmTechAI. Taken in their totality, they represent a snapshot of the state of AI from researchers, manufacturers, software developers, scientists, and professors. Understanding AI is fundamental to understanding how it will impact your work and personal life. The final statement, from Microsoft Research, transcends mere optimism and promises revelations that can radically change our world. Let’s buckle up.
Mark Lien, LC, LEED AP, is industry relations consultant for the IES.
1 R. Williams, “OpenAI released research into how using ChatGPT affects people’s emotional well-being,” MIT Technology Review, Mar. 25, 2025.
