Why we are betting on ACI, not AI

Adam Cunningham, Chief Strategy Officer, 02.27.23

02.27.23 Adam Cunningham, Chief Strategy Officer

When it seems the entire world is doubling down on Artificial Intelligence (AI), we’re doubling down on Artificial Collective Intelligence (ACI). We’re not simply being obstinate, we actually think we’re onto something that matters. Before you write us off, bear with me while I explain our thinking. Still unsure? Run the little experiment at the end. Sorry ahead of time for any forthcoming heartache.

It’s just adding one word at a time. As Stephen Wolfram so perfectly summarises in his wildly intelligent (but wildly complex) article about how ChatGPT works, it’s simply adding one word at a time - albeit, rather intelligently. We’ll save you the trouble of reading the piece, but said with a bit more detail, think of it in three simple elements:

  • ChatGPT writes by using probability to produce a “reasonable continuation” of whatever text it’s got so far.
  • To produce a more interesting essay, ChatGPT sometimes randomly picks lower-ranked words.
  • The probabilities for each word come from scanning billions of pages of human-written text.

Simplified this way, it’s quite easy to see how AI technology, like ChatGPT, can appear helpful, but perhaps cause more issues in the long run if not given guidance and not sense-checked. This is also why a lot of AI-generated text can read and sound very “flat”.

So, while ChatGPT, a form of AI, on its own, doesn’t solve everything, let alone fully replace jobs, it (and Large Language Models in general) can accelerate certain graduate-level tasks while producing a high-quality output. Examples might include:

  • Creating short or long summaries from documents
  • Brainstorming different concepts and giving idea directions
  • Performing high-level concept analysis across multiple documents or research
  • Conducting deeper dive analysis based on the above analyses

These tasks, normally given to humans, would certainly take far longer to complete. AI systems get them done quickly and efficiently, and at a quality-level worth recognizing. But the humans aren’t out of a job just yet.

Let’s get the acronyms out of the way.

Okay, before we go deeper, let’s make sure we’re on the same page with terms.

  • ChatGPT is just one element of AI called LLM, or Large Language Models.
  • AGI, or Artificial General Intelligence, refers to the development of Al systems that can perform any intellectual task that a human can. AGI systems are designed to exhibit general cognitive abilities and human-like intelligence, not just to support human decision-making in specific domains. AGI is often considered the ultimate goal of Al research, and is a highly ambitious and long-term project.

We at AGM think LLM is promising but flawed on its own and AGI, while also interesting, is years and years away (and will probably still require humans). This leads us to:

ACI, or Augmented Collective Intelligence, refers to the integration of artificial intelligence with human intelligence to enhance the collective intelligence of a group of individuals or a network. It involves the use of Al to support human decision-making, collaboration, and knowledge management, among other tasks. ACI systems are designed to augment human intelligence and work in partnership with people.

Simply put: we’re supercharging our humans.

Putting LLM-based tools in the hands of an intelligent human supercharges them: quantity of work and quality of work improves.

This is why we’re betting on ACI. AGM + AI = ACI.

The deployment of LLM and AI to our teams globally not only democratises certain cognitive tasks but also accelerates our ability to iterate. Go with us on this one:

  1. Imagine all the problems you could solve with an abacus.
  2. Now someone gives you a calculator. It still requires you to use it, but now you’re starting much further along and solving bigger problems faster.
  3. Now imagine you’re given Excel with formulas. You’re still needed, but the starting point is incredibly further along and you’re able to solve far more complex questions.
  4. Now you’re given AI. You’re very much still required, and in fact, arguably more important to the equation, as your environment, experience and expertise make the AI useful.

Still unconvinced ACI is the better bet? Run your own experiment.

Cool, no problem. You already have Apple’s LLM at your fingertips (and you don’t use it).

  1. Go into your Messages app on your iPhone.
  2. Find your most important relationship.
  3. Open up your chat with them and tap the first word Siri suggests you type above the keyboard.
  4. Then repeat 100 more times.
  5. Hit send.

Point is: you might use AI to assist, edit or expand, but you don’t use it on its own.

We won’t either. Environment, experience and expertise still matter.

In the weeks ahead, we’ll be moving past the philosophical to breakdown how we’re deploying ACI across our services and solutions globally. It’s pretty cool - namely because our humans are still involved and getting supercharged by the minute.

Have a pressing question? Get in touch here.

Find out what we can do for you