There's a backlash brewing against AI.
It tramples our rights. It harms our environment. It erases us. People are angry and they’re finding outlets for that anger. Just a whiff of AI in a communication or on social media can create a viral smear campaign, destroying years of hard-earned goodwill. This puts mission-based organizations in a tricky spot.
We all have more work than we have time. AI lets us do in hours what would otherwise take days. That means more time for the mission. But when that mission exists to uplift humanity, it’s hard to justify using a technology that so often diminishes it.
That got us wondering: is it possible to use AI and honor your values?
To answer that question, we need to understand what AI is and why it’s become so controversial.
What is AI anyway?
AI is just math. It doesn’t “think” in any real sense of the word. You give it some text, it does a bunch of calculations, and it gives you text back. Those calculations happen in a mathematical model called a large-language model, or “LLM.” The LLM is what we think of when we say “AI.”
You can think of an LLM like a mathematical equation:
Your text + LLM = AI response
LLMs differ from familiar operations like addition and multiplication in one important way: they don’t always give the same result.
In this way, using an LLM is a bit like baking a loaf of bread.
Your ingredients + oven = loaf of bread
Even if you use the same ingredients every time, each loaf will be slightly different. Maybe the oven is running warmer today. Maybe you didn’t leave it in quite as long. Maybe it's a bit more humid. Whatever the reason, you’ll always get a slightly different result, but not so different as to be unrecognizable.
You'll still get a loaf of bread. It's not going to turn into a fruit salad.
LLMs are similar. If you ask an LLM the same question twice, you'll get slightly different responses. The responses will be similar: ask a question about the solar system and you'll get an answer about the solar system. But the way that answer is phrased and the information that's included will change based on the path the LLM followed to get from your question to its response.
All of this makes it feel like you're having a conversation with a thinking entity, but you really aren't. Just as your oven isn’t doing anything conscious when it bakes your bread, an LLM isn’t doing anything conscious when it answers your question. It's all math. So what’s so controversial about math?
Why the backlash?
We gave three reasons for the backlash against AI at the start of this post:
- It tramples our rights
- It harms our environment
- It erases us
Let’s look at each of these claims in a little more detail.
AI tramples our rights
The “Large” in “Large Language Model” is an understatement.
LLMs are massive. They’re built upon the indiscriminate collection of public information from the internet. The collection methods often disregard copyright, trademark, and privacy rights. The companies behind the major LLMs obscure how they use that information and who they share it with.
People object to the use of their publicly shared information for the private profit of these companies. They especially object to its being used for mass surveillance. So more and more people now go out of their way to keep their data away from AI models. Some are putting their sites behind logins, while others are taking them down entirely.
Mission-based organizations fight for our rights. AI vendors trample them.
AI harms our environment
Massive models demand massive resources. It takes a lot of power to do all that math on all that data. The big AI companies are all scrambling to build dedicated facilities, called datacenters, to store all that data and do all that math.
AI datacenters create a huge drain on local power and water supplies. They generate biological and electronic waste on unprecedented scales. They're also loud and unsightly (think motorcycles running next to your house all day).
Datacenters have existed for decades. What makes AI datacenters different is their scale. As an example, Utah was just forced to scale back plans for a datacenter that was going to be three times the size of Manhattan. A victory of sorts for the local community, but even the scaled back plan is 1.5 times the size of Manhattan. Who wants that in their backyard? Nobody.
People fight AI datacenters everywhere they're proposed. That, in turn, has led AI companies to try to hide the projects or move them to marginalized communities that are less able to resist them.
Mission-based organizations defend the vulnerable. AI vendors exploit them.
AI erases us
The promise of AI was that it would eliminate the dull, mechanical elements of our lives and free us for creative and intellectual pursuits. But that's not what has happened.
Instead, AI is most often used to simply eliminate jobs in cynical cost-cutting moves. Everywhere you look, you find companies laying off staff or withholding pay in the name of “investing in AI.”
When it doesn't eliminate jobs, it frequently replaces the creative work. We’re inundated with an undifferentiated mass of mediocre writing, art, and software. You can sense this every time you read something and think “this sounds like AI.” There's even a term for it: slop.
Mission-based organizations dignify work. AI vendors eliminate it.
It doesn't have to be this way.
A more ethical AI
It’s no wonder that people are revolting against AI. At every turn, we've used it to degrade humanity.
Now that we understand this, we can start to envision an AI that uplifts humanity. Unsurprisingly, it's the opposite of everything we just described.
Ethical AI respects our rights
An ethical AI has to respect ownership and privacy rights. It can’t do that if it’s built by copying the entire internet. It has to use less data.
Fortunately, LLMs aren’t the only game in town. There are also small language models, which are exactly what the name implies. They use smaller datasets and less math to do the same thing that LLMs do. This means they’re less capable than the mammoth LLMs, but they’re powerful enough for most purposes.
Ethical AI preserves our environment
No technology that competes with people for natural resources is compatible with a value system based on human dignity. This is especially true in a time of widespread environmental upheaval. For our AI to be ethical, it has to have modest power requirements.
Small language models don’t need anywhere near as much power as LLMs. Many of them can even run on your laptop: I’ve been running some on my 5-year-old Macbook for a few months. I do have to be thoughtful about how I use them. But I haven’t had to steal my neighbors’ electricity, and so far nobody has complained about the noise (though I do sometimes play my music … enthusiastically).
Ethical AI amplifies us
Technology is at its best when it amplifies us. AI is too often used to make us irrelevant instead. An ethical AI would make us more valuable, not less.
There’s plenty of work for AI that doesn’t displace people. It can be hard to see that because we so often focus on what comes out of AI: writing, images, code. But what if we think about what goes in instead?
AI understands language.
This has two important implications.
First, it makes technology easier to use. You don’t have to bother with finding the right file, knowing which buttons to click, or remembering arcane commands. When you want an LLM to do something, you just ask and the LLM does it. How much time have you wasted trying to learn a new program or remembering how to do something on your computer? With LLMs, you only need to know what you want to do. It’s the LLM’s job to know how to do it.
Second, it lets computers handle paperwork. There’s no end to paperwork. This is work that requires linguistic skill, but not creativity: taking notes, writing proposals, filing reports. It's necessary work. Without it, we lose the funding and support that are vital to our mission. But any time we spend on paperwork is time we aren't spending on the mission. LLMs let us turn the paperwork over to our computers.
An ethical AI takes on the tedious tasks that sap our spirits. It frees us to do the meaningful work that serves our missions and nourishes our souls.
Human-First AI
We've been giving all of this a lot of thought, and we're beginning to form a different vision for AI. We call it “Human-First AI.” We're still figuring it out, but it's something like AI that:
- respects human rights (no mass data collection, no surveillance)
- runs at human scales (your laptop, not a small city)
- makes computers more approachable (interact with them on your terms, not theirs)
- liberates people instead of replacing them (do the mechanical work, not the creative work)
We don’t think AI should force you to choose between your mission and your values. By selecting small models to target tedium, we can build AI that puts people first.
We're tired of technology that demeans humanity.
Human-first AI is technology that celebrates humanity.
That's what we're building. We can't wait to share it with you.
