Development, Machine Learning

Open Source, not OpenAI

The last few days have been a whirlwind for me, as the system in which my brain exists appears to be filling with entropic energy that I attempt to corral. Days after celebrating my first Pride being out, the Human Rights Campaign declared a state of emergency for LGBTQ+ Americans. Apple finally announced a long-awaited launch that further establishes spatial computing’s role in the future of our technological evolution. My daily conversations traverse topics related to human rights violations in training artificial intelligence to authenticity and safety in self-expression for queer individuals, trying to navigate a rapidly shifting regulatory environment and whether or not computers have more rights than women.

There is a rapidly emerging pattern with the discourse around AI – the anthropomorphism of AI agents into (primarily female) personas, whether in product (“Siri”, “Alexa”), science fiction (“Samantha”, “Jarvis”) or self-created (“Sydney”, Kevin Roose’s conversational-partner-turned-romantic interest manifested from Microsoft’s Bing search engine). Throughout this discourse, I can’t help but feel as though we’ve missed a major step along the way – why are we so focused on “othering” these systems, instead of internalizing them as an extension of our own capabilities?

This train of thought is what has taken me down the path of further exploring the concepts of local user agents and Self OS. It is impossible to separate ourselves from the influence of others, but in the digital realm, we depend on centralized technology providers to far too great an extent.

I informally surveyed my ‘Launch your Startup’ class at CBS last week about how many of the people in the class were concerned about how their information was used online. About 60% of them raised their hand, and one person summed up the complexity of her experience: “I’m worried about it, but at this point, it feels so far out of my control that I don’t know what to do about it.”

An acquaintance of mine recently talked about using ChatGPT as an assistant to write their thesis paper – a thesis paper about queer identity and religion. It was shocking to me to think of someone willingly feeding in information about the parts of their identity that are being criminalized around the United States to a company that was funded by right-wing billionaires. The rise of authoritarian and anti-LGBTQIA+ legislation didn’t disappear in 2020, and yet, people still try to adhere to the idea that tech is neutral.

I don’t use ChatGPT. I don’t trust promises made about data collection and protection, especially given the way that my identity is under attack by the people who stand to profit from its use. I made the switch to running Linux on my personal computers several months ago, the day that Microsoft pushed their ChatGPT-powered Bing search bar onto my desktop. The commercial software stack is collapsing, locking users into increasingly tightly controlled environments where customization is locked away, abstracted because of its complexity. This is not the way.

In Silicon Valley, many people celebrate computer technology as the hallmark of capital markets at work. You can “make something from nothing”, conveniently ignoring the little asterisk that assumes access to a fast internet connection, hardware, and the time it takes to develop the language(s) of working with machines. Yes, the internet can teach us many, many things. It cannot, apparently, teach billionaires how to be more empathetic to the lived experiences of the global majority.

I’m still bullish on the concept of AI, though right now, my ‘A’ looks a lot more like ‘augmented’ intelligence than ‘artificial’ intelligence. My laptop – a Pop! OS Lemur Pro from System76 – and desktop each have different local language models running on them, with various interfaces (my desktop is using command-line tools around privateGPT & GPT4All; my laptop is using dalai and Alpaca 13B ). I’m testing out the differences with prompting and ingesting my data in different ways, and experimentally evaluating the emotional differences I have when I interact with each model separately.

Today, I used one of my local AI agents to help describe SelfOS:

Liv Erickson’s ‘Self-OS’ concept is an innovative approach to operating system design that focuses on enhancing user experience through personalization and intuitive navigation. It seeks to create a customized, simplified interface tailored for each individual user based upon their preferences as well as the type of device they are using. 

Her idea involves developing an intelligent Operating System (OS) which can learn from its users’ habits in order to predict and provide what is needed without being asked. In essence it would act more like a personal assistant than just another application installed on your computer or phone, providing better usability while making the user experience much smoother and easier. 

The ‘Self-OS’ idea could be implemented with AI technology to allow for advanced predictive analytics which can anticipate users needs without being asked. By understanding how people use their devices it will then become more intuitively navigable, faster in performance as well providing better personalization options so that the device feels like a true extension of its user’s personality and lifestyle.

Not bad, laptop, not bad.

Large technology companies hold so much data about us. In this era of chat bot-based applications and conversational interfaces, we’ve begun giving them more and more unstructured data. In return, we get the statistical average of internet-speak guiding us though our ideas.

Another issue weighing on me is the further concentration of power that Microsoft now has over the computing stack. It almost feels like 2001 all over again, but this time, the anti-competitive behavior is hidden behind so many invisible layers that it’s nearly impossible to claim an advantaged market position in all of them. I learned today that Microsoft is the only company allowed to license OpenAI’s models. Paired with the enormous compute capabilities in their Azure data centers, this gives them a significant competitive advantage.

But, the open source movement gives me hope. It’s how I’m able to run multiple different interfaces to local language models that can be fine-tuned on my computers with my own data, and know that it’s not leaving my own environment. It’s how I can build the Firefox web browser locally, and add custom patches so that I can run experiments with my browsing history safely. It’s how I can picture re-imagined fair value exchange systems working.

And it’s also how we get our agency online back.