Responding to Generative AI
The headlines are rife with Artificial Intelligence updates. Some of the articles spell out concerns and fears. Justified concerns if you have ever seen any number of films set in a dystopian future. Others celebrate our inevitable march towards an all knowing Artificial Intelligence (AI). Most miss the nuance of what is general AI, compared to the demonstrated practical applications to date. All the exhibitions remain far from Artificial General Intelligence.
Read "A Citizen’s Guide to Artificial Intelligence,” to understand the difference between machine learning and artificial intelligence. I also suggest taking an introductory data science course. These classes offer a glimpse of the staggering effort required to create a predictive model for a specific topic area. They also show what it would take to predict across all human knowledge on a subset of that history.
This distinction is very important. When we use the term AI like this we evoke the wrong ideas on how far we’ve come and also how far we have to go. Also, we underestimate the need to bake in safe guards because we mistake the application to date completely benign. The popular applications out there are some of the most complex and rich machine learning algorithms. Employed in specific domains such as computer generated art, writing, gaming, process automation, autonomous driving, and similar. What we have not seen is Artificial General Intelligence in its truest sense. Think of Data from Star Trek beginning as a child and learning as he develops. Grappling with the complexities of morality processed through black and white logical rules.
The first thing we could do to contain the conversation is to be specific on what we are doing and not doing. We are not, in the very near future, creating a sentient and autonomous mind born out of code that teaches itself along the way and makes its own decisions on which thread to follow and an understanding of the consequences that come with those decisions. Outcomes with scopes that range from program failure to Judgement Day. That said, we are putting the fear of God into the hearts of creators and workers at large everywhere. One of the critical aspects to being human is creativity. Something that for many of us we express through our work. Creative work has generally been safe from industrialization and automation. Though Warhol and others employed those industries in their craft.
“To be creative is to tap into our mental pool of resources, our experiences, and our connections and use them to create something new, revised, practical, or outlandish. Human creativity is our ability to move back and forth between the realms of ‘what is’ and ‘what could be.’” – Agustin Fuentes
Our life stories are punctuated by our creative outputs. These exclamation points mark the moments where we tap into our full humanity. Moments where we enter a flow state and touch our full potential as Maslow would describe it; "What a man can be, he must be." If generative AI takes away avenues of our full expression as human beings, then what outlets remain for expressing our purpose of being–consumption? I hope not and pray those same alarm bells go off in everyone else. That said, we should not reject machine generated content and assistive AI given the positive impacts they could have. Some examples of a balanced application include:
Wikipedia entries generated by tools like Chat GPT paired with AI generated images from Adobe Sensei and proofread by humans.
Product and Services FAQ’s generated by tools like Chat GPT paired with AI generated images from Adobe Sensei and proofread by humans.
Faster Proof of Concept beta validation via automated computer testing.
These applications are pragmatic and less invasive and can leverage generative AI now. Regardless, when we see our creative contributions shrunk down to output from code, we can’t help but feel our humanity challenged. Along with feeling our incomes and purpose cut down. There are valid arguments for how these generative machine models can aid the creative process. We are in the nascency of these AI capabilities. Like most paradigm shifting achievements, the landscape operates like the Wild West. In this type of setting, we count the bodies and assess the damage only after plundering the gold mines. We can get ahead of the damage by leveraging another disruptive technology.
Blockchain, in layman’s terms, is a digital record keeper of transactions. What differentiates it is the security around it and the distributed nature of the record keeping. Peer-to-peer (P2P) networks of computers (users) that can read and validate every transaction in the record book are the backbone of the ledger. To date, the blockchain networks have been both proprietary and sudo-privately managed. The next evolution is to create a network that leverages modern user profile creation and management tools like AppleID along with public authorities. All as users on this P2P network can create, validate, and store a digital rights & authorship record book of content.
Blockchain while going through its own growing pains, works as a viable technology by Cryptocurrencies and NFTs. It could serve as the global registers of AI Generated content; also of human created digital content. This approach preserves the dignity in human work by authenticating the authorship of digital content. Software helps produce machine generated deep fakes, art, writing, and more. Software can also help to digitally stamp (think watermark) what is machine generated from what is human crafted.
Every time a model generates content, it gets registered in a global block chain. A service managed by governments and searchable by teachers, news media, and others. Assuring us of what is real, what is fake, and what is additive but not a human original. Anything unsigned or registered gets categorized as UNKOWN. The key is the participation and leadership from the major players like Apple, Google, Adobe, OpenAI, Microsoft, and governments.
Boston Dynamics has a role to play as well. They are producing some of the potential physical machinery and software combinations that could change the face robot / human interaction. It is high time they and others bake in Asimov’s Three Laws of Robotics into every model, robot, and experiment. William MacAskill has been a pioneer in championing longtermism. If we are to protect the quality of life and sense of purpose for future generations in a world with sentient machines, we need to make sure those automatons don’t kill us the second they objectively observe our morally conflicted history in detail. That said, AI could be a meaningful lift for mankind helping us unlock solutions for long vexing problems if we hardwire in protections from the start.
