Ah, the singularity. What does that mean? And does it have any impact on someone weighing how to embrace AI as a supporting technology system for their efforts?
People ask me questions about AI’s future and about the cybersecurity, societal, and long-term employment risks all the time as part of our adoption conversations. If people are asking privately in one-on-ones, then others are surely pondering similar issues.
Before I dive into these four topics, let me state generally that unless the specific topic directly impacts your company and its business, it’s probably a distraction. Most businesses have a very hard time implementing AI, in large part because adoption often centers on a frontier model that isn't built for a specific type of business. These large one-size-does-all generative AI tools require quite a lot of internal or professional services work to make them impactful.
There are incredible amounts of money invested in AI, both privately and in public markets. As such, companies and their CEOs leverage the media to drive interest and belief in the future of their technologies. In doing so, they often cite the potential promises and impacts of their models.
Several instances have illustrated a lack of discipline when it comes to PR, most notably promising the near arrival of sentient AI systems or artificial general intelligence (AGI), and the widespread replacement of white-collar workforces. This has boomeranged on AI companies over time, tarnishing their reputation and creating a general public dislike for AI.
So much of the news cycle around AI, while dynamic and often sensational, is frequently speculative and has very little to do with actual use. Keep your eye on the prize: Determining if and how tools can help you achieve your goals.
The singularity refers to the arrival of sentient AI. It, in essence, argues for recursive self-improvement, intelligence explosion, and loss of predictability. It’s largely an argument or position propagated by OpenAI CEO Sam Altman, most recently after the Hugging Face GPT 5.6 Sol cybersecurity hacking incident.
Of note, the AI accessed Hugging Face without human direction. That directly impacts our next topic: Cybersecurity.
However, most experts view Altman and other AI enthusiasts’ almost worship of the singularity as the resurrection of the AGI myth. Claims of achieving sentient AGI were debunked over and over again, causing tech execs to back off their proclamations. Unlike AGI, you cannot measure the singularity as it is a process so it can be argued for, much like Gods and spirits can be debated.
That being said, most importantly, the singularity has not arrived. If AI were self-improving, it would certainly address its probabilistic decision-making and hallucination challenges first thing. So it does not impact you. It is not near.
For tech companies like OpenAI, discussions of the singularity provide air cover for responsibility and a sense of omnipotent power for their technologies. As you can imagine, such spin impacts market investments, culpability, and pricing.
A lot of companies and individuals want to know how secure their data is with an LLM. Given we have reports from both OpenAI and Anthropic of autonomous hacks by AI agents, are companies at risk? Currently, no. Those models have not been released to the public and are not in use. One would assume they would not be released until these matters have been resolved.
Generally speaking, it is as safe as any SaaS platform if you are using a private account and have turned off the LLM’s ability to train on your data. Not using a private account is a mistake, and you should assume all of your data is being used by the company that owns the platform.
This response may not be comforting at all if you know much about cybersecurity. Importing corporate data from other platforms, particularly with not well-designed MCPs, can add even further risk to the equation. However, if you are using tools like two-factor authentication and are authenticating access to data regularly, you are taking steps to protect yourself.
On a macro scale, yeah, there is a lot to be concerned about. Current development and deployment methods for frontier models have created higher risk. While you see some members of the AI community calling for slowdowns and token gestures to embrace regulation, tech companies will keep developing as fast as possible to maintain or achieve market advantage. That’s how the system works. External regulation is necessary to prevent unbridled technology development.
Current regulatory approaches to AI should not inspire confidence, either. Haphazard blocking of model exports so far seems to be last-minute, uninformed stopgaps. Unfortunately, government officials are as uninformed about AI impacts as the general public, often reacting to public concerns.
Concerned persons should wait to implement new frontier model releases for several weeks in case there are major issues. This is no different than any other new product. And please put pressure on your elected representatives to implement smart AI regulation. It may not be much, but it is something.
From impacts on the energy grid and the environment to the destruction of societal fabric on many levels, there are great concerns about AI’s secondary consequences. I don’t have good answers for them.
And, I am not going to apologize or gloss over these very real concerns. I was candid about this in my conclusion to Now Is Gone, which you can also read here (first draft). Like so many of us, I feel powerless about the AI tsunami. Having been through a technology wave or four, I know fighting professional use of new tools is a pointless exercise. Implementing them mindfully, however, is a professional choice each of us can make.
In addition, I think we are looking at an eventual sea change in the type of models we use, as I noted in my last article. Frontier models are incredibly inefficient and costly on so many levels. For AI to be more successful on qualitative and cost grounds, models need to be narrower and tailored to specific industries and use cases.
On a larger level, the government must resolve these impacts, as it is apparent that industry will not self-regulate. Unfortunately, as noted in the cybersecurity section, this is going to take a groundswell to resolve. While data centers have become politicized, continued pressure on other AI-related issues needs to be applied at every level of government.
Finally, there is the issue that touches almost every single one of us. And that is workplace impacts. While the tech industry's AI layoffs have continued, there are increasingly conflicting data points showing some companies are having to hire back employees, walk back cuts, and others are simply hiring more people as a result of using AI.
What is clear is that AI will impact almost every job in some way, simply because it is the new software of the current moment and long-term future. I sense most people understand that AI is not going anywhere, and as much as some may dislike it, it has become unavoidable.
Embracing AI as a technology tool that can help perform tasks is the right course, I believe. It’s the one I have taken, and candidly, the best way to influence the future is to be a part of it.
One thing is certain: What we believe AI is today will change. Now is gone. As AI evolves and moves beyond the current frontier model era of department store AIs that do everything, job roles will evolve, too. I think we have much change ahead of us, and a lot of that is unforeseeable.
Some of it has become obvious, including strategic and qualitative oversight of AI technologies. Some of the coming change will surprise us. Whatever happens, resisting it will not help us achieve success. Walk through it.
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.