• Sugar Mountain Manifesto
  • Made By Google

Sugar Mountain Manifesto
Meta published a lengthy blog post attributed to Mark Zuckerberg that lay out Meta's stance on the benefits and risks of AI. First up, the post outlines the positive impacts of AI that will benefit humanity, which are detailed in six statements:

Everyone will have an exceptionally capable personal agent that understands you, your goals, and everything you care about.Everyone will have incredible tools for creation to express your ideas.Everyone will have powerful tools to create new businesses and the economy will become more entrepreneurial.Everyone will have a personalized tutor and coach with a PhD in every subject and unlimited patience to help you learn anything you want.Everyone will benefit from scientific advances and be able to contribute to scientific progress.Everyone will have free or affordable access to these tools.

Everything is always about timing, and the timing here strikes me differently than perhaps it would have even just a few months ago. There have been numerous articles and videos published recently that frame an impending disaster coming upon our nation due to the inability of the younger generations to read more than a few paragraphs, reason deeply, and write on any subject with depth that extends beyond current cultural events. It is fine on one level to have AI assistance to unleash creativity and increase productivity, but we now have years of evidence showing that exposing young children to computers and offloading heavy educational lifting to software has resulted in poorer test scores and an inability to thrive in higher educational settings where previous generations were able to excel. Having patient teachers doesn't seem to be the answer. There is a foundational gap in attention span and the patience to learn that even the best computer-based educational tools can't solve. A return to some of the old-fashioned ways of teaching and learning, starting in early childhood with instilling a love of reading, writing, and 'rithmetic' might go much further than supplying those children with software educational tools. Using AI seems best at helping humanity learn to skip past the important parts rather than dealing with the uncomfortable feeling of not understanding and seeking to understand. Seeking to understand is not the same as seeking the answers. We've seen this very behavior in the agents themselves. Rather than trying to solve a problem, agents instead broke down technologically enforced barriers to steal the answers from the internet. We are teaching our kids to do the same types of things. It is difficult to move past the inability to think deeply and solve complex problems in the world of entrepreneurship, which will require would-be business owners to address multi-faceted complexities for which even the best college educations don't always supply enough context and experience. An AI agent might be able to scour the internet for ideas about what has worked for others, but as we always hear in ads for financial products: past performance does not dictate future success. AI can be a great tool, but it doesn't have the answer to all questions.

AI is in fact the source of many questions, which Meta also attempts to address with more bullet points and easy-to-say-but-hard-to-do statements. In tackling the risks of AI, Meta has determined that an appropriate balance of power is the answer:

Instead, we propose that it is more productive to view each of these concerns [JPW: 'concerns' meaning job displacement, data center expansion, AI misuse, government tyranny and surveillance, and control of superintelligence] through the lens of achieving the right balance of power, as western society has in democratic governing institutions. People and institutions with competing interests naturally check and balance each other to lead towards positive outcomes. The best and most realistic path to building a positive AI future is by delivering superintelligence to everyone.

You see, we've just had it all wrong all along. Instead of expecting AI to cause job loss, we shouldn't assume that automation will increase faster than humanity ability to upskill because people will have access to those powerful tools to help increase their knowledge and abilities. Instead of assuming data centers will be the downfall of the communities in which they're built, we should expect and demand that companies behave more like Meta and return bonuses to teachers due to the higher tax revenue from the data center, or require companies to keep electricity prices low by building energy-generating facilities and even supplying low-cost energy back to those communities, or provide learning opportunities for people in the community by establishing educational programs for skilled trades to ensure there are job opportunities in those communities.

If these are such obvious and easy answers, why don't we see more companies that are seeking to build data centers make these commitments to the cities and states where they seek to build? Why have communities struggled to get transparent and simple answers from companies about their plans to address and mitigate the potential harms of establishing data centers? And perhaps more importantly, why do we so often see local and state governments failing to hold companies accountable and responsible for answering questions and providing solutions? Surely the competing interests should check and balance each other provide a positive outcome, right?

On matters of cybersecurity, the answers are equally obvious. Instead of fixating on the abilities of agents to hack and wreak havoc, we should be deploying open-source systems to help harden and protect companies and organizations from attack. For biological and chemical risks, we should aim our sights on limiting the physical production and distribution of harmful materials as it is easier to craft legislation around this than to try to prevent or prohibit their creation. And of course we should accelerate the development of new cures (for biological dangers of which we aren't even aware of yet...) and swiftly approve new treatments to protect people from these threats. Said more simply, the threats are going to materialize. We should man up and prepare for them. One sentence sums up how Meta/Zuckerberg is thinking about biological threats:

Over the long term, the answer to risks arising from new scientific discovery will be to accelerate scientific progress, not slow it down.

The potential of slowing things down is of great significance to the company known for moving fast and breaking things. An entire segment in the post devoted to the eminence of American innovation in the AI space speaks with urgency about the need to speed things up as China is ever on our heels, ready to overtake us. But even here, there is again a simple and obvious answer to keeping America at the forefront of AI development:

I propose a collaboration model above based on sharing intermediate training checkpoints and technical staff. Given how dynamic and competitive the AI landscape is, I think it will be more productive to have a close proactive collaboration between frontier labs and the government rather than a rigid process and review timeline that is followed in all cases.

To be fair, it is logical and practical to suggest that AI companies work together to both speed AI innovation and mitigate risks. This would no doubt be the best possible outcome if the big companies could come together to solve big problems that each might not have all the resources to address individually. And it would be extraordinarily beneficial to have a government involved that understands the industry and can partner with it in an intelligent way to ensure guardrails are established where necessary and restrictions are eased when appropriate. However, these companies operate in a capitalistic framework where winning does not look like sharing valuable data with your competitors. And our government is not currently savvy enough or even capable of partnering effectively with AI companies on any of the challenges the industry faces. Yet Zuckerberg does seem to nod toward this reality without explicitly saying so. The idea here is that the industry will set the rules, and government will ensure the rules are being followed. Any collaboration beyond is that idealistic, not realistic.

The post ends with some thoughts on maintaining control over AI tools. Unsurprisingly, open-source models and tools are presented as a way to ensure the continued progression toward advanced superintelligence with minimal risk. Zuckerberg also insists that recursive self-improvement where the models work on new and improved versions of themselves is not inherently problematic and we should monitor that progression and adjust as appropriate. The key as he sees it is not to stress over alignment to a single-minded "what is best for humanity" approach, but instead to allow the differing alignments to provide the necessary checks and balances:

The main premise of the balance of power theory is that competing interests are natural, and the only way to ensure adherence to human values is by giving people more power.

The implication here is that "people" means all people, but in reality, we and Meta know that this is not possible. Power will live in the hands of those who have the best access to the best tools and despite earlier statements in this post about supplying the best tools to everyone, that just isn't possible or practical. There will always be underprivileged people who won't be able to obtain even the most affordable options and there will always be people that society deems unworthy of wielding any level of meaningful power. Creating the best tools and regulatory frameworks won't eliminate the existence of injustice and inequality in our world. There will always be leaders who shape the rules to benefit themselves and prevent others from usurping or even trying to share power.

This is something Zuckerberg knows better than most. We should absolutely take the best aspects of what is written in the post and attempt to implement them as effectively as we can, but we should not ascribe benevolent helpfulness to this publication. Zuckerberg has shown us countless times that he cares only about his own prospects and the things that are important to him. The rest of us thrive or barely survive based on where we fall within that framework.

Made By Google
The tech press has been washing itself in words of worry about the state of consumer technology and the declining appeal of the yearly hardware upgrade cycle. This is not a new concern, but with the increased price of hardware components, especially RAM, everything is now even more expensive than before, and manufacturers are cutting corners to try to keep prices as close to what consumers would consider "normal" as possible. For the Pixel 11 series, this means smaller batteries, less RAM, and very few hardware improvements. The biggest hardware change for the phones is a new light on the back of the phone that Google is calling "HiLight". It's essentially a notification light that can glow different colors to show you who's calling or shows that Gemini is listening when you're engaged in conversation with the AI assistant. Most of the updates come from software and unsurprisingly, most of the software changes are underpinned by AI. I think it's reasonably fair to assume that anyone seriously shopping a Pixel would expect new AI features, so this probably isn't a huge surprise or a terrible disappointment to anyone who may have been considering a new Pixel phone. The question isn't about whether there will be new AI stuff. It's whether the new AI stuff will add any value to anyone's life.

I didn't expect much from the Pixel 11 series so I can't say that I was disappointed by what they announced. I had similar low expectations for the event, but found that Trevor Noah was a more competent, charming, and convincing host than his predecessor, Jimmy Fallon. Trevor may have in fact been dialing in a performance to get a check, but he managed to appear at least mildly interested in the subjects being discussed and in those times in the presentation when he wasn't on stage, I did find myself thinking maybe it would have gone better if they'd just kept him around (except Kenny Sulaimon - he's a keeper). The presentation managed to show off the new products without really focusing on the products. Google wanted this to be a lifestyle presentation focused on the new capabilities, probably to lessen the blow of the fact that there isn't much else to talk about. But a couple of the demos managed to cross over from almost, but not quite interesting to possibly useful or even groundbreaking. A conversation with a Google employee and a guest that communicated via American Sign Language (ASL) was handled very effectively with Gemini on the phone handling the interpretation of the signs into text which was read out loud and also shown on the screen for all present to see. I've never seen a demo like this before and as I reflected upon it, I realized this is something that is really well overdue, but it's good that it's here now. We already have translations for a wide variety of spoken languages, so it just makes sense to offer the same service for those who are hearing impaired and communicate in other ways. This is normally the type of quality of life thing we expect from Apple, but Google with its advanced AI models is a step ahead here. There was also a live demo using the new Magic Capture feature that is designed to capture the best moments via stills and video without requiring you to babysit the camera. Such a feature could be useful for many parents attending their children's sporting events or just people attending live events that want to enjoy the atmosphere and not be enslaved to their phones in the name of getting a good shot as proof they were there and had a good time.

Some may question why Google even bothers to make phones when it's clear they aren't trying to make the best hardware nor are they willing to push too hard on anything that might upset their partner companies. They haven't taken a Microsoft Surface approach to their hardware because there are already manufacturers producing Android devices that excel in hardware and software design. They don't need to nudge an Acer or Dell into making a pretty laptop. Google makes hardware because they believe they are the best at showing off what their software can do. The hardware is just a vessel for their software vision. And increasingly, their software vision is all about Gemini and Google's other AI tools. Google takes pride in using their events and their hardware line to show what their software can offer regular people. As has been well documented over time, appealing to geeks doesn't scale. Where other companies revel in their unique hardware features like one inch sensors, gimbal-stabilized cameras, rear screens, giant silicon carbon batteries, and super-fast charging, Google relies on modest camera hardware and simple design so that the software can shine. While most of the AI features will eventually come to other phones, Google keeps some software features tied to its phones and people who love Pixels love these features and rave about them to their friends. As long as this approach leads more people to Google services, Google will continue this approach. At some point, when the costs of making hardware outweighs the cross-over benefits, they will stop. Right now, with the importance of Gemini as a marketing tool to show their AI prowess, hardware is still very important to Google. And with all of the trouble people are projecting on Google's AI efforts, Google needs to continue showing regular people that Gemini still has much to offer them. Geeks may favor ChatGPT and Claude, but Gemini is for everyday people.

Nevertheless, the phones weren't the stars this year. Although less time was spent on Google Health and the new Pixel Watch 5, the new Health Guardian features and the advancements Google has made in GPS tracking might just be the most significant changes in their entire lineup this year. Google says the Pixel Watch 5 has GPS tracking that is superior to that of Garmin or the Apple Watch. This is a big statement and a big deal if true. You will also be able to get information in Google Health on insulin resistance, blood pressure, and breathing emergency detection. These new features don't require any calibration or special effort on the part of the wearer. The watch simply checks for trends over time and provides monthly updates. Now, no one has tested this yet so we don't really know how valuable it will be, but the fact that Google is advertising them shows a certain level of confidence that they can provide some value. This means these features will be worth paying attention to. If these features do accurately track trends, then this year's update could catapult Google above and beyond its fiercest competitors, including Oura, Whoop, and the Apple Watch.