- Auto Matt Ick
- What is AI Safety?
- A Spectacle, a Snap Judgment
Auto Matt Ick
You have undoubtedly been impacted by WordPress even if you've never heard of the company or used one of its products. About 40% of the internet consists of websites built on WordPress and WordPress holds a share of the content management system market that nears 60%. WordPress.org was started in 2003 by Matt Mullenweg and Mike Little. WordPress itself is an open-source project, but in 2005, Mullenweg started Automattic to support the growing ecosystem around that project. Automattic is responsible for WordPress.com, which provides hosting and content management services, as well as WooCommerce, the plug-in system allowing users to add features to their websites such as stores, subscriptions, and memberships. Mullenweg has been its CEO since 2014. But on September 9, he was placed on paid leave by the company's board. We still don't know why this decision was made, nor do we understand why it seemed to happen so quickly. There was a faint scent in the air of eau de OpenAI/Sam Altman as he returned to his job by Friday and the board members that spearheaded the vote to remove him exited the company. The public still has no idea what happened; no one seems to be talking, but we did learn that during the 30+ hours Mullenweg was on leave, the CFO (who immediately became interim CEO) and the chief legal officer signed severance agreements for themselves that became effective on September 10 and provided each with 12 months of salary, one year of medical coverage, accelerated vesting for their equity, and the ability to exercise their vested stock options. Since Mullenweg fired them upon his return, these agreements now seem applicable. However, since the conditions under which they were created are somewhat dubious, there is an effort being made to determine if they are in fact legally binding.
TechCrunch has a good summary of the most obvious possibilities for Mullenweg's ouster:
Sarah Perez
These events can be interpreted in two very different ways. In one scenario, Automattic’s board is responding to an internal leadership crisis, voted to put Mullenweg on leave, and then established protections for the executives who could face adverse consequences if the intervention failed (as it now has) ...
In another scenario, it could appear that Automattic’s board was trying to create a window of control for some other reason — perhaps a strategic transaction — by putting Mullenweg on leave and taking the reins. According to sources, this is the theory Mullenweg suspects — though by his own account, the board never gave him a stated reason for the vote to begin with, leaving him, like outside observers, largely to speculate about the underlying motive. That absence of explanation, combined with the CFO’s stock sale, appears to have fed his suspicions and factored into his decision to retake the CEO role and remove the board and other executives.
I don't typically follow or pay attention to the removal of CEOs from major companies unless it is a company that interests me or the CEO was a personality of which I was aware. I asked AI if it is normal for CEOs to be removed in the manner as we saw with Mullenweg and apparently, it is rather common. Here are the primary reasons given for not providing the public with reasons for dismissal:
What it can signal
The absence of a stated reason is not proof of scandal. A CEO change may stem from:A performance or strategy disagreement.Loss of board confidence.Succession planning or a negotiated exit.Health, family, or personal matters.A governance dispute.An ongoing investigation where premature public detail would be inappropriate.
That said, an abrupt move—particularly if the board appoints an interim CEO immediately, limits the departing CEO’s access or authority, or uses unusually formal language—usually means the board viewed a prompt transition as necessary. It can be a serious situation, but outsiders generally cannot responsibly infer the specific reason without evidence.
I would say in this case we can likely eliminate 4 of these scenarios - performance, succession planning, health, and governance. That leaves us with loss of board confidence and an ongoing investigation. The TechCrunch article I quoted suggests the possibility of the active lawsuit with WP Engine and Mullenweg's potentially questionable actions around destroying evidence may be a reason for seeking to put him aside. The fact that the CFO and CLO are the ones that led the charge to push him out may be indicative of concerns around the legal proceedings as opposed to just general uncertainty about Mullenweg's leadership and trustworthiness. The fact that they wrote up severance agreements for themselves signals that perhaps they did not ultimately believe they would prevail and this was a way to protect their future exit from the company or this was some kind of collusion between them to facilitate a comfortable departure, perhaps because they haven't been fans of the direction of the lawsuit or the company itself. As Mullenweg himself noted, the fact that the CFO sold his shares in the company before any of this transpired may be a clue.
I don't know much about Mullenweg other than what has been written about him and his various struggles with his company and partner-turned-adversary companies like WP Engine, but it will be interesting to see how this continues to develop and whether he can continue to maintain control of the company he started.
What Is AI Safety?
For the record, I do not believe that AI will be the thing to destroy all humanity. I think humans are more than capable of this job. However, I do think that the conversations around making AI safer are perhaps misguided from the start. I'm not an AI expert and I don't have any expertise on training large language models. But if it is safe to assume that LLMs are trained on human communications, then misalignment is built in and is guaranteed. As every human being is flawed, so also are the things we say, write, and do. We talk of cheating, sneaking, stealing, conspiring, obfuscating, and all manner of deception, usually in the aim of accomplishing some goal. I think science fiction misled us into thinking that artificial intelligence could be built upon the best and highest quality knowledge in our universe and would therefore reflect the best moral realities. And indeed, AI is built upon the best knowledge we have available to us, but it is also built upon the worst and lowest quality human activities. We have read that the internet was scraped and scoured to obtain training data. We have heard of the complaints and lawsuits from musicians, graphic artists, writers, and just regular people on Reddit that their content and creations were used to train AI models with no compensation or even acknowledgement from the AI labs.
We aren't surprised when babies and young children mimic our good and bad behavior because we know they model what they see and hear. We work to try to train them to understand why bad behavior is bad so that they won't do it. This is in fact a lifelong process. We never get to the stage where we completely eliminate all behavior that would harm others or ourselves. So if we haven't been able to fully address misalignment in humans, what would make us think we could address it in anything we create? But it seems flippant at best and misanthropic at worst to suggest that we need to determine what our tolerance is for AI misbehavior. Maybe there are some assignments or jobs where our tolerance is so low that we would not want AI to be involved in any capacity. And maybe there are other tasks where the risk is low enough that it could be doled out to an AI model or agent without much, if any, supervision. For commercially available AI tools, it might be possible to enforce sufficient guardrails to keep tools reasonably safe for mass market consumption.
The nearly impossible scenario to defend against is the nefarious or simply foolish use of advanced AI tools in the hands of users who aren't bound by licensing agreements or regulatory governance. There is no real way to stop hackers and attackers from using AI tools to their best advantage. The best approach we have is to be as prepared as possible for those attacks and use AI tools to protect and defend. The cybersecurity wars will continue. Real wars will as well. The stories of this week have shown us that it is becoming increasingly difficult to manage the issues of alignment when training and testing models and in the United States, the government has shown a complete unwillingness and inability to do anything to intervene, and because the perceived financial incentives are too great for the AI labs to ignore, we can realistically expect that things will continue in the general direction in which they have been going. It will be up to companies and individuals to determine how we want to live with AI.
To be clear, I do not think we should stop trying to train AI tools in an effort to mitigate potential harms and damages. I just do not think that we should expect those efforts to be completely successful. And we need to start dealing honestly with what that means for society and our future.
A Spectacle, a Snap Judgment
Snap has now allowed some tech influencers and journalists to try the $2195 Spectacles augmented reality glasses. Snap is aiming these glasses more toward Apple's Vision Pro than to Meta's $799 full color smart glasses, thus the significantly higher price tag. Unlike display glasses, there is no dangling cable and puck containing the computing power and unlike most smart glasses, the glasses don't require a connection to a phone to display information. The device has its own built-in computer, and you can call up the home screen by looking at your palm. The glasses have hand tracking, so you use your hands to navigate the user interface. The glasses will have a small selection of apps available at launch, including Spotify and Shopify. Streaming services such as YouTube, and HBO Max will be accessible via the built-in web browser. Navigation services and live translation will be available as well. One unique feature of the glasses is the ability to play games live using the glasses with another user that is also wearing a pair of the glasses. Sure, this is an expensive gaming option, but for anyone who planned to have more than one pair in their household, it means some experiences can be shared, which is not the case for the current crop of smart glasses, display glasses, and AR/XR headsets. The glasses will also feature an AI assistant that will surface useful information for you. The assistant will also be available for iOS and MacOS. One of the biggest obstacles beyond price is the battery life: they are rated for four hours of mixed use, which means that you will not be able to get a full workday out of them. Evan Spiegel, CEO of Snap, says that he expects the glasses to be worn in shorter spurts and not constantly or for many hours at a time. The company has partnerships lined up with Salesforce, Nvidia, and AWS and a cellular capable model will be offered by Verizon for $2395 plus a $10/month service fee for connectivity. Interested parties will also be able to try out the glasses at select Verizon locations. Snap plans to open a pop-up store in Los Angeles and other locations in the future as well.
The Spectacles are very expensive and very awkward for many face shapes so it's unlikely that they will be a commercial success. Right now, more people would prefer to spend $2000+ on a folding phone than on a pair of glasses. Nevertheless, this concept of glasses with their own computing power and connectivity, plus the ability to broadcast information and entertainment in your field of view without necessarily blocking the view of the world in front of you is likely the future of smart face wearables. Eventually products like these will become more compact, battery life and computing power will continue to improve, prices will come down, and we will see future iterations of products like these become more mainstream and useful to a larger market. In the meantime, these will probably remain a small niche used primarily in the enterprise and industrial markets.
Member discussion: