Today, I wanted to write a quick post about something I continue seeing and getting asked about. Earlier this week, I was quoted on Dark Reading that I agreed with the NCSC that cybersecurity threats from ChatGPT were overhyped. So I get asked, if ChatGPT won’t supercharge attackers, then why do I keep seeing article after article to the contrary? There’s a simple explanation, what you are seeing is parroting.
I saw this article yesterday. Analysts Share 8 ChatGPT Security Predictions for 2023. Let me save you the trouble of digging into the article. All of the predictions about lowering the barrier for criminals and supercharging attackers aren’t based on any reality. There is no evidence for any of this. If that were true, we’d already see a meaningful impact right now, but we aren’t. These predictions are written by people who’ve never used ChatGPT in any of the contexts they are making predictions in. People are just repeating other people who’ve also never used ChatGPT for this purpose. With hot topics come hot takes. But this creates a strange false consensus, similar to a filter bubble on social media.
When people think something has become a matter of consensus, psychologists have found, they tend not only to go along, but to internalize that sentiment as their own.
-Max Fischer, The Chaos Machine
There are also a lot of reports of conversations about ChatGPT on the Dark Web. Here is an example. Reading articles like this, you are meant to believe that criminals’ use of these tools is accelerating, but there’s another explanation. Just like all of us are talking about ChatGPT and exploring the surface of capabilities, it makes sense that criminals would as well. I’m just applying Occam’s Razor here. Talking about a topic on a forum isn’t industrializing that something in attacks. I haven’t dug into the content of any of these posts, but my gut tells me they are experiments and not any kind of industrialized real-world attacks. That doesn’t mean attackers aren’t riding on the hype of ChatGPT to launch attacks. We’ve seen this play out with fake ChatGPT tools with embedded malware.
That Doesn’t Mean These Tools Aren’t Useful
You can say that you believe in machine learning and deep learning’s ability to help professionals address real cybersecurity challenges (which I wholeheartedly believe) and still be critical of the hype surrounding tools like ChatGPT. There are certainly advantages to using these tools. When it comes to LLMs, people seem to discover that more mundane aspects, such as text summarization, text generation, constraining to a knowledge base, etc., help solve business problems and increase efficiency. People may not realize that LLMs have been pretty good at these tasks before ChatGPT came along.
Overall, I think we should be prepared to be surprised. People find interesting and unexpected ways to use technology. That’s the great part about human ingenuity. Just be mindful that because a lot of reporting seems to be saying the same thing, that doesn’t make something true. Controlled lab experiments aren’t the real world and any accuracy and usage statistics from providers should be taken with a grain of salt. Think critically about reporting on this topic, even my own. Nobody made me the Chief AI Whisperer. I’m judging my perspective looking at the realities, capabilities, and limitations of these tools, not filling in the blanks and shouting from the rooftops like some crypto bro saying how ChatGPT is going to 10x everything.
I have a longer post that goes more in-depth on some of these issues and misconceptions for next week. For example, there are real cybersecurity threat from these tools, but it’s in their integration into applications without consideration of the attack surface. More to come.
On Valentines Day this year, I commented about the current AI boom and the paperclip maximizer, but I feel this topic deserves a bit more explanation.
With all of the hype and mobilization caused by the ChatGPT demo, the warning about AI alignment is missing from news coverage and conversation. We are getting a preview of the playbook for an even more advanced AI, and the plays are unsettling. These plays aren’t unexpected and are the same ones people concerned about alignment and safety have warned about. The same plays that AI maximalists downplay as people overblowing the dangers or not understanding how development works. Well, there’s quite a bit of vindication in the alignment and safety crowd. However, I’m not sure they are happy about it.
If a more capable AI system is developed and deployed this way, humanity is in trouble, possibly headed for the paperclip factory.
This post uses recent events to dismantle three fundamental tenants AI maximalists use to downplay dangers. We look at the conditions and activities surrounding the release of ChatGPT and consider the actions compared to a more capable AI system, even systems such as Artificial General Intelligence (AGI) and Artificial Super Intelligence (ASI).
Definitions
AI Alignment
AI Alignment is the area of research that aims to have the goals and activities of an AI system match the intended goals of the designer of the AI system. This seems like a fairly trivial problem, but as we’ll see with the paperclip maximizer, it’s not so simple.
Consider HAL 9000 from 2001: A Space Odyssey for an example of an alignment problem.
2001: A Space Odyssey HAL 9000 Scene
Also, for a great short story on a misaligned AI system, check out Philip K. Dick’s Autofac. Please read the story, and don’t watch the garbage Amazon remake they did for Electric Dreams.
AI Alignment and AI Safety are two different research areas, but for the sake of this blog post, I’ve lumped them together since this topic concerns both areas.
The paperclip maximizer
For those unfamiliar, the paperclip maximizer is a thought experiment that shows how a seemingly aligned goal applied to an advanced AI can have unintended consequences. In this case, AI is given the goal of maximizing paperclip production. In service of this goal, the AI converts everything in its environment, including humans, into materials for maximizing paperclip production.
The goal is to show that problems with an advanced AI could be far worse and have devastating impacts on humanity. That’s the frame of the concerns in this blog post.
Simple Formula
There is a simple formula that AI researchers use to discount the concerns of alignment and safety folks.
Responsible Development + Appropriate Guardrails = AI Utopia
Given the stakes, they argue that companies developing such technology will do so responsibly and with humanity’s best intentions in mind. Also, appropriate guardrails will be in place if something goes wrong, protecting humanity from harm. These points crumble in the current playbook with LLM-based chat technology like ChatGPT. We are getting neither responsible development nor appropriate guardrails.
Responsible Development
I wouldn’t classify what we’ve seen in the past few months as responsible development. We’ve seen a technology with many known issues thrown in the production environments for which they are not suited in the hopes that the issues can be corrected in use. It’s the old analogy of building an airplane while you are flying it.
This makes sense for these companies since incentives are aligned with making money, not making humanity a better place. ChatGPT made cutting-edge language models available to regular users but wasn’t any leap forward in innovation compared to competitors, it was just a great demo. Some would say a publicity stunt.
Some instructive takeaways from the ChatGPT release are informative to the release of a more advanced system. There is a rush to develop, release, and compete. This isn’t a revelation but a preview of a playbook we’ll see with a more capable technology, where responsible development is needed even more. Many tech companies now see themselves in an AI arms race, which won’t lead to more responsible decisions regarding releasing new technology.
So, what did we learn from the release of ChatGPT that would be instructive applied to AGI?
A company with a demo having known issues can build a lot of hype by exposing it. This hype forces adjacent companies, who previously acted more cautiously, to release their demo to be seen as “competing” and not be left behind. With caution now to the wind, there’s a rush to push it into production systems, even without the necessary safety protections in place, with the mindset that minimal protections are a good start and more advanced protections can be bolted on after the fact. Now, imagine that what these companies are building is HAL 9000’s
As we can imagine, a rush to public demo, even in the face of known issues may be the default starting point for future AI releases.
I’m all in favor of for-profit companies, but there needs to be some consideration for potential harm from irresponsible release. We wouldn’t let a biotech company engineer and release new viruses at will. The stakes are also much higher for a more capable AI, such as an AGI or ASI system. Today’s AI systems can and do certainly create harm, but the size and scale of the harm increase for a more capable system.
Guardrails
Guardrails are another fundamental tenant of keeping an advanced AI aligned with our goals. One of my favorites is people saying, “If it starts doing things we don’t like, we’ll just unplug it.” As if that would even be an option. “I’m sorry, Dave. I’m afraid I can’t do that.”
The guardrails around the current crop of LLM-based chat systems aren’t working out very well either. These systems begin life completely open, exposed to the words of humanity, and happily provide you with what you ask for. You know, things like how to hotwire a car, blackmail someone, and will even make stuff up for you. This behavior is not what the system designers intended, resulting in a patchwork of attempts at trapping these conditions.
The problem is that the space of all potential unwanted behavior is vast and unknown at deployment time resulting in issues that aren’t known until you see them
The problem is that the space of all potential unwanted behavior is vast and unknown at deployment time resulting in issues that aren’t known until you see them, meaning you can’t ideate traps until after the condition presents itself. There’s no reason to think it won’t be even worse for an AGI system with a far larger problem space, it will depend on techniques and methods that haven’t been developed yet, but the problem space will still be vast and challenging for creating guardrails.
Of course, this also makes a huge assumption that it’s even possible to develop appropriate guardrails for all of these conditions in the first place. We’ve just begun to scratch the surface because we assume accidental conditions. And then come the purposeful attacks.
It seems there is much in the way of misunderstanding in the attack surface of these systems. Although a system like a chatbot may seem pretty simple and have a single interface (text), you end up with something reasonably complex from an attack surface perspective. A system with a single interface but with limitless numbers of undocumented protocols, all waiting to be exploited. And, surprise, you find all of this out after you’ve deployed it into the world.
A system with a single interface but with limitless numbers of undocumented protocols, all waiting to be exploited.
Any advanced AI system that requires deployment into the world to develop guardrails will not end well for humanity. The bad news is it’s very likely that an AGI system would also start in this completely open condition, as we see in current transformer models, requiring appropriate guardrails to keep it aligned.
It was obvious to these companies that the guardrails around transformer-based chatbots were insufficient, but these companies moved forward anyway. This should be a wake-up call.
Job Displacement
Let me take a slight detour here to make another point on the current environment of generative AI. One of the other claims from AI maximalists is that people shouldn’t worry about losing their jobs because we’ll figure out how to share the wealth with everyone affected. This has always been a laughable proposition, but we are starting to see how laughable it is.
There is no reason to think that an organization with an AI capable of performing a job will have any consideration for how it affects the humans previously doing that job. I’ve written about AI taking both hobbies and jobs Current generative AI companies including Facebook are not sharing the wealth with freelance writers and artists they are displacing or about to displace, and there’s no reason to think this will change in the future. Job displacement must be addressed by other means, including possible government intervention, policy, and taxation.
Conclusion
In conclusion, I think we got lucky. We blew through a stop sign, and no cars were in the intersection. Many guardrail bypasses we’ve seen haven’t led to any realized harm. We wouldn’t be so lucky with a more advanced system, so it’s time to see this as the warning it was.
Whether we want them or not, chatbots are coming to search engines. Google announced Bard and Microsoft implemented ChatGPT style integration with GPT-4 from OpenAI in Bing. Well, at least people are talking about Bing again, which hasn’t been mentioned in a conversation since its launch.
All jokes aside, this begs the question, is search really ripe for disruption? Many in the tech world seem to think so or at least hope so. My guess is it could be. For a long time, the optimization around search has been to put more eyes on more ads. The question is, is the current crop of chatbots the answer?
Implementations
Just like the answer to most questions regarding the use of new technology, it depends on the implementation. I haven’t really thought about search for quite some time and as a researcher, I admit that my usage of search engines may differ quite a bit from an average user. But, logically, it seems that when people use search engines, they are looking for answers to simple questions or information on current events.
If you look at the trending searches on something like Bing, all of these are current events.
When it comes to information about current events, it will be interesting to see what, if any, impact there is from chat in this space. The current crop of chatbots like ChatGPT were trained on data from 2021. Obviously, a breaking news story won’t be in its memory.
The search engines also seem to be exposing additional functionality in much the same way the ChatGPT demo did, which does open up some interesting possibilities and problems. For example, when people look up medical information and get disinformation or advice in general and get bad advice. People may create different relationships with search engine sites than they have previously, and this isn’t always possible to anticipate ahead of time.
Just the facts, occasionally.
Sometimes you just want the answer to a question. What’s the circumference of the earth? What’s the Tom Hanks movie where he made a wish and transformed from a kid into an adult? Was William Shakespeare a real person? In these cases, it would seem that a chatbot with accurate information could be beneficial and preferential to traditional search. Even in more complex scenarios where you just want an answer, you won’t be bogged down by additional content or need to search different sites to confirm information. This would be great in a perfect world.
But…
We all know the current crop of language models tends to hallucinate facts. Examples of this are all over the web. How this is handled will largely dictate whether these features are useful or useless. I personally don’t think we are anywhere near having a universal knowledge machine that falls within the bounds of acceptable failure and that seems to be playing out in this scenario as Alphabet loses $100 billion dollars as their shares tank after Bard hallucinated facts.
Guardrails
If we conduct a thought experiment where we have a perfect knowledge system, there are still plenty of problems to think about. Consider more controversial subjects or even topics that haven’t been settled. How will these chatbots handle these situations? How will it boil a complex topic down to a simple response? There’s been a heavy-handed approach to the guardrails on these topics, so much so that it borders on the absurd. It’s important as a society that we address uncomfortable topics and come to a resolution. It’s not like we’ve settled everything there is to settle and we can all go home.
My guess is that heavy-handed guardrails will remain in place in the spirit of “safety.” As a matter of fact, it’s already happening as Microsoft’s chatbot refuses to write a cover letter for someone saying it would be unfair to other candidates. So, the most useful features may be out of reach of the people who need them most.
Unfortunately, this filtering will lead to a lot of technology blame as people lead with their biases. We’ve seen this game play out on social media.
Nudges and Forgetfulness
The obvious failures will continue to be identified in the system, but what about the not-so-obvious failures? What happens when something disappears and we don’t notice? We seem to think historical events are seared into our consciousness or as though the Internet doesn’t forget, but it does. Imagine what would happen if Google suddenly stopped returning responses for a historical event. Would that event, in effect, disappear? Probably, for a great many people anyway, depending on factors such as the cultural importance of the event and its age from memory.
This removal of events could be purposeful, such as Chinese search engines not returning results for the famous “Tank Man” image in Tiananmen Square. Baidu even took it a step further and made sure their image-generating software, a competitor to DALL-E, wouldn’t generate a version of the photo either.
What’s even scarier than any purposeful manipulation is when the system accidentally forgets things.
What’s even scarier than any purposeful manipulation is when the system accidentally forgets things. Events could disappear and be lost from collective memory. Equally as bad is that the system hallucinates “facts” around an event, changing its focus. What would the implications be if this happened to a world event as impactful as The Holocaust? Maybe it didn’t disappear completely but the facts around the event changed. What if fifty years from now, the memorials around the event were the only reminder? What if we started to question why we built them in the first place? This absolutely terrifies me and it should terrify you.
For my previous example, I chose the most extreme case to illustrate a point. It’s unlikely that an event such as The Holocaust would be lost to history, but getting the facts wrong is certainly possible. Often the facts shape the impression of the event. This is a conspiracy theorist’s dream.
We need to be careful in the rush to implement a technology, that we don’t lose or manipulate things we actually care about. Could these systems nudge us into forgetting or not caring the way we should? The answer is yes. Will they? That remains to be seen.
New Perverse Incentives
Chat-based knowledge systems make quite a bit of sense for search engine companies as they are data collectors. There’s an old saying that goes something like this, “People lie to their friends, but nobody lies to Google.” Well, nobody is going to lie to a chatbot either, and collecting this information is going to be a gold mine of personal data. This is a power-up to the current crop of perverse incentives.
I made the following joke on Social Media.
[chatbot enabled search engine] What’s the best burger place in town?
[response] Billy’s Burger Joint has the highest rating. Oh, by the way, did you know you could save hundreds by switching to GEICO?
It’s a joke with some truth to it. There’s no way to ad-block that.
There will certainly be hands on the scale nudging people in particular directions. Sure, search engines can manipulate results today, but delivering an answer in the form of a knowledge base holds more of a user’s attention and gives it appears to have both consensus and authority. Imagine giving bad advice, like throwing batteries in the ocean, or if the GEICO commercial was much more sneaky and was part of a recommendation.
Bad For Privacy
Regardless of perspective on usefulness, we can all agree that chatbot integration in search engines is very bad for privacy. This is in an environment where additional logging and analysis will have far greater scrutiny by human eyes, at least in the short term, due to analyzing and improving the implementation. Richer forms of personal data and even thoughts will be collected, stored, and analyzed with the intent of weaponizing this information.
We can all agree that chatbot integration in search engines is very bad for privacy
Are we getting features we didn’t ask for?
I don’t necessarily blame the tech companies here. Sometimes with innovation, you don’t know you want something until you see it. That’s what Silicon Valley is hoping for. Only time will tell if these features truly enhance the search experience, is useless, or turns it into a nightmare. It seems to be leaning toward the latter at the moment.
Regardless of getting things we didn’t ask for in search, it’s certainly coming for other products. 2023 is going to bring a lot of things we didn’t ask for because, currently, these companies are justifying their investments.
Microsoft announced that it was bringing ChatGPT features to Microsoft Teams. This is a head shaker for me. I need to restart Teams up to five times a day due to issues, but now it’s getting even more features. Microsoft Teams now resembles the big box store PC you get loaded up with bloatware.
Microsoft Teams now resembles the big box store PC you get loaded up with bloatware
All of this rush to implement technology we know has issues in production systems reminds me of an Arthur C. Clarke short story called Superiority. In the story, a technologically superior planet is defeated by another one because of their willingness to discard old technology without having perfected the new. Just another example of how fiction can inform reality. In the future, will fiction be our reality?
On a daily basis, I’m bombarded by fantastical news stories and email articles about ChatGPT and how 2023 is the year it’s going to replace “X” job or profession. It seems that no profession is safe. It reminds me of the Dewie The Bear scene from the movie Semi Pro where Will Ferrell screams, “Everybody panic” in a crowded coliseum.
ChatGPT has gone from being a piece of technology to a social contagion right in front of everyone’s eyes. Claims are being made that people can’t possibly believe, but they are making them anyway. I wrote last year about how ChatGPT creates amateur futurists. Well, it seems the floodgates are open.
This has created a group of overoptimistic zealots who’ve basically taken on the identity of crypto bros pumping ChatGPT to 10x. On the flip side, I’m certainly not delusional about the job loss myself. It’s true that generative tools will certainly have an impact on jobs. To take this further, I also expect some surprises and unexpected cases to crop up. The world is a complex place and it’s just not possible to take in all the variables. There are lots of people working in this area, and automation is going to have an impact on multiple professions, even if it doesn’t eliminate them. So, fair enough. But, the extent of the current mania is mind-boggling, with people tossing all manner of delusional predictions at the wall, hoping one sticks.
So, I wanted to add a bit of sanity, so anyone can make sense of this topic and at least try to frame these news stories in the appropriate bucket.
Evaluating News About ChatGPT and Job Loss
There’s a simple way to evaluate the merit of these job loss arguments, even if you aren’t familiar with the technology. Just ask a simple question. Is the cost of failure low? Boom, that’s it. If the cost of failure is low, then there’s a good chance there’s a near-future risk of impact from these tools. If the cost of failure is moderate or high, then there’s little chance of impact in the near term with the current crop of AI tools.
In my previous post, I wrote about how freelance artists will be impacted by the pervasiveness and accessibility of these tools. If you don’t like the art generated by the tool, just generate another one. On the flip side, think about the impact of having ChatGPT be your doctor or lawyer, especially given the fact that these models tend to hallucinate facts.
I know ChatGPT passed the Bar exam, but why is this surprising? I mean, I think most people would be able to pass the bar given an open book and unlimited time. Knowing the answers to questions is a far different matter than applying the knowledge you have to specific situations. I certainly wouldn’t want ChatGPT to argue my case in court, even though creating convincing BS seems to be one of its strong suits.
A Warning
I’ve stated before my biggest concern with all of the ChatGPT hysteria is that people in their mad scramble to compete will end up using this technology in areas where the cost of failure is not low, where there’s the potential for harm and even loss of life.
I worry quite a bit about mental health uses and the medical field in general. Mental health chatbots should be a bit red flag for us. There have been promising uses in using computer vision models to assist doctors in identifying whether tumors are malignant or benign, but that’s far different than having a knowledge system like ChatGPT. Even if these tools reach the status of “pretty good” it will lead to an automation bias where the doctor takes the recommendation of the system by default. This condition would, in effect, make something like ChatGPT, your doctor. Would this be better or worse than clicking through WebMD and diagnosing yourself?
I think people, in their optimism, tend to fill in the blanks, even in the case of very complex problems. In such cases, assuming we are only a couple of tweaks away from solving existing issues. It’s dangerous to bestow some mystical object status on technologies such as ChatGPT when what we need is a realistic analysis of the capabilities and limitations of such approaches. We tend to underestimate the complexities of even simple problems, which makes humans terrible at predictions, but our over-optimism could lead us down a very dangerous path in the next couple of years as these tools creep into critical decision paths.
2023 is going to be an interesting year for generative AI. Past the demos of the previous year, there’s going to be a big push on monetization. This application and integration into products will have a displacing effect and this effect will have a disproportionate impact on certain areas, mainly in the creative arts.
Note: All images in this post were generated with the prompt, “generate a stock photo image that is just okay and mildly pleasing to a human“
Perspective Mismatch
I see quite a few people making statements like, “AI isn’t going to replace people. People who use AI will replace people who don’t.” Or some form of this statement. This is usually followed by a hearty high-five at their insight. The problem with this statement is it just doesn’t reflect reality for certain types of jobs. To put this in another context, this is like saying, “Food service robots won’t replace food service workers. Food service workers who don’t use robots will be replaced.”
These comments are often made by people in positions that have no visibility into impacted areas. If you are a developer who finds a productivity boost from CoPilot or ChatGPT, you have a much different perspective than the millions of independent artists, creatives, and copyrighters across the globe. The warning signs are here. This isn’t AI hate or denialism. This is reality.
2023 will be the year where you start seeing freelance creatives losing opportunities as the impact of generative AI hits freelancers and the gig economy. Jobs will be eliminated or devalued to the point where it isn’t worth it.
From Demo to Application
Last year was the year of the demo, the time to showcase the novelty. Demos like Dall-E, Stable Diffusion, and ChatGPT were everywhere. To quote the philosopher Steven Pearcy, “Nobody rides for free.” In 2023 the novelty is wearing off and the check comes due. These systems cost money, and the need to monetize is here.
This year will witness a rapid expansion of generative AI in all areas with a low cost of failure. Everything gets an AI assistant. This assistant mindset will put the tools much closer to everyone. You won’t need to go to a website or use an API to use the tools. They’ll be built into the things you use on a daily basis. Microsoft is investing billions of dollars into OpenAI, and comments from Yann LeCun from Meta about generative models being built directly into Facebook for creative activities. Notion is beta testing a generative writing assistant integrated into the tool. And the list goes on and on.
ChatGPT was a great demo, but not really overly unique. Other large tech companies have similar tools but are more conservative about how they use and release them. Well, the floodgates are open.
Good Enough is Fine
Creative arts are an area ripe for disruption and democratization, aka devaluation and elimination. In the areas of creative arts, the impact of failure for an AI system is low, making it a great target for AI, which tends to have high rates of failure. This is much different than, say, a self-driving car not recognizing a red light or a pedestrian in a crosswalk. In creative areas, If you don’t like something, just generate another one, or ten, or a hundred.
Human novelty won’t cut it for most regular creative tasks. Sure, some creatives will continue to find work, but the work won’t be the same and there’ll be far less to go around. Even though humans have the potential to bring something unique and ultimately deliver a better product, with the devaluation of artistic endeavors, good enough will be just fine.
This technology will most likely be justified under the heading of “advanced customization.” Being able to take something like an advertisement and custom tailor the content to specific groups or even possibly down to individuals. Today, you can target ads at specific groups of people, but in the future, you may not need to. By having some form of adaptive ad with adaptive content, the ad can tailor itself to the individual. Don’t get me wrong, this is an interesting technical problem but it creates a human problem. Once again, my tech friends are reducing everything down to the process of doing something and not the human aspect of the task. Whenever you have extreme customization and add scale, you always create a process with too much friction for human effort.
This is about the time that I hear the groaning about these creatives having the same access to tools that everyone else has. This is certainly true, but it’s not an equalizer at all. Even if creatives were able to keep pace, there’s a monumental satisfaction drop. If you are an artist who enjoys putting together graphics, having an AI generate a bunch of them isn’t the same as creating your own. If you are a VoiceOver artist, having an AI generate a voice isn’t the same as using your own voice. This isn’t the “eliminate the mundane work so humans can focus on more interesting things” promise from AI being fulfilled here.
I get the feeling in discussions on this topic that technologists equate creative writing to writing code.
I get the feeling in discussions on this topic that technologists equate creative writing to writing code. This is incredibly disconnected. Creative endeavors aren’t some problem to be solved or some slog that needs to be eliminated. In many cases, the slog is kind of the point. As someone who does both, I totally get it. It’s completely inefficient and hard to understand, but it’s the reality.
But this is irrelevant because creatives won’t be able to keep up, especially with these tools at everyone’s fingertips, tempting people to do it themselves. Many just trying to get tasks done will use the tools provided to them in the interface. They’ll wonder why they are paying someone to push the same buttons they can push right in front of them.
Why am I so confident? We have similar evidence for this trend in camera phones. Look at what camera phones and photo filters did to the photography industry. In the past, people would hire a photographer or have a lab touch up photos, but now people use an iPhone and smear a filter on it. This is true even for events as meaningful as their wedding. Sure, we still have photographers today, but the industry is pretty decimated. Many who could have made a living can’t even eke out a meager side hustle with this.
When I was growing up, I used to hear the term “starving artists.” I guess it’s time for creatives to realize that existence again. Even Fiverr will have to rebrand to Freeverr.
Near Future
They’ll notice their work first change, reduce in value, and then drop off.
Most immediately impacted will be freelance artists, graphic designers, writers, and copyrighters. They’ll notice their work first change, reduce in value, and then drop off. Below are a few examples of what I see happening very soon.
Freelance copyrighters, instead of being asked to write the copy themselves, will be asked to tweak and make adjustments copy generated from a tool. They’ll be asked to do this work for far less than they’d be paid to write the copy. In many cases, the copy will be tweaked in-house and the external copyrighter won’t be contacted.
Say you are part of some group that wants to have some t-shirts made. Instead of having an artist design the image and graphics for the shirt, you now use the web interface of the t-shirt company’s website to automatically generate unique artwork for you, bypassing the need for a freelance artist.
Artists are asked to become part of a working group where they receive a gift card for providing feedback on a swath of images generated by an AI for a stock photo site. The mechanical mashing of thumbs up and thumbs down is a stark contrast to the artistic activities to which they are accustomed. Less pay and less fulfillment.
What Do We Do?
So, what can be done about this? Nothing. It’s beautiful in its simplicity. There’s literally nothing we can do. Genies don’t fit back into bottles. I could say that companies should put an emphasis on hiring people to continue doing this work, but it’s not going to happen. And, maybe there’s nothing we should do. This may just be the natural part of human evolution toward becoming machines ourselves. If you look at some of the jobs impacted, these are uniquely human jobs.
If I have hope here, it’s that we begin to tire of everything being automatically generated and tailored specifically for us.
So, let’s pour one out for some of the jobs impacted in the near future. Here are a few off the top of my head, artists, graphic designers, stock photography, VoiceOver and voice actors, copyrighters, proofreaders, and a whole lot more. Coming soon to this space actors and actresses.
A couple of years ago, I said that the people who should really be worried about deepfakes are actors and actresses. Interesting thought, since many of these generative AI systems have NSFW protections built into them, pornography will continue to be made by humans. The future prospects of human actors and actresses look quite a bit different through this lens.
Who Really Loses?
Is humanity destined for a future where “creativity” is defined by just smashing the like button on the output of AI?
My biggest concern here isn’t the job loss, it’s what we are losing as humans. As I mentioned, these are jobs that were previously uniquely human. 65,000 years ago, our ancestors showcased their creative talents by painting on cave walls. They may have been memorializing events, but it’s a sure thing that some of them liked the activity and it gave them a sense of fulfillment.
Is humanity destined for a future where “creativity” is defined by just smashing the like button on the output of AI? Only time will tell.
Scour the web and social media for reactions to OpenAI’s ChatGPT and you’ll notice a trend. Everyone’s now a futurist. The tech crystal ball has revealed to droves of people, regardless of background and experience a future where you can ask for technical content in the style of Shakespeare or even write college essays for you. People are crawling out of the woodwork, making wild predictions based on nothing more than the fact that of their surprise at the output. It conjures images of Tom Smykowski’s Jump To Conclusions Mat.
So, why does everyone believe they can now peer into the future after interacting with ChatGPT?
State of the Art
Let me say that the team at OpenAI did some fantastic work here. ChatGPT is legitimately cool research, and OpenAI is pushing the state-of-the-art forward and doing it in a public forum where people can evaluate the results. I can’t wait to see what the next iteration looks like. None of my observations is a reflection on the work they’ve done.
Manufacturing Futurists
There are two primary reasons ChatGPT turns people into futurists. The first is their surprise at the output, and the second is the accessibility of the demo. The first is fueled by the second.
Most people playing with the demo have never interacted with a Large Language Model (LLM) before. This lack of interaction means they don’t have a baseline for comparison of progress. A majority of the responses of the system will therefore be surprising. Surprise, in some cases, causes people to ignore the genuine failures in others.
The genius (or intelligence) in ChatGPT lies in the accessibility of its demo. Everyone can access and can play with the demo with no knowledge of programming, and it is all delivered on a simple web page. No complex decisions to make and no parameters to tune, just a blank input and your imagination. You have to love the simplicity.
I’ve seen wild claims that ChatGPT may be AGI or at least is Proto AGI. This is nonsense and ignores the system’s genuine failures when generalizing to the real world. I’m not an AGI researcher, but I can tell you we won’t get to AGI by chaining a series of LLMs together. Although, this should be a warning because if this is what AGI looks like, humanity is pretty screwed.
GPT-3.5 Isn’t A Single Model
The first thing to remember is that GPT-3.5, which is behind ChatGPT, isn’t a single model but a series of models.
Models referred to as “GPT 3.5”
GPT-3.5 series is a series of models that was trained on a blend of text and code from before Q4 2021. The following models are in the GPT-3.5 series:
code-davinci-002 is a base model, so good for pure code-completion tasks
text-davinci-002 is an InstructGPT model based on code-davinci-002
text-davinci-003 is an improvement on text-davinci-002
This explains why ChatGPT can be both good at Shakespeare and Python programming.
Over Optimism
Over-optimism can lead to early adoption and implementation inside products, which can lead to devastating consequences. What people forget is that this is just a demo. Many of the surprises people had at the capabilities of ChatGPT are because of their questions. Even though the questions may seem complex, they are often in the confines of a narrow problem definition. Humans aren’t good at randomness or complexity. We often oversimplify scenarios and don’t account for real-world complexities.
“Give me a recipe for tomato soup in the style of Shakespeare.” We become in awe of the prose and ignore the quality of the recipe, which in this case, is technically what you were asking the system for.
Surprise leads us to gloss over the many, many failures that ChatGPT has. It even fails on simple tasks that revolve around keeping count of items, like this example from Elias Ruokanen.
Even in the domain of information security, I saw people raving about ChatGPTs capabilities in identifying vulnerabilities in code. Someone commented that it could be used to collect bug bounties in the cryptocurrency space, but it failed. In my own experiments, I often found that it misclassified issues making guesses, such as since a parameter was accepted, it was vulnerable to SQL Injection.
Speaking of code understanding, since ChatGPT could provide explanations of code, people thought they could use it to answer questions on Stack Overflow, with the predictable banning that followed.
In experimental systems like ChatGPT, Dall-E, etc., the cost of failure is next to zero. You don’t get that in the real world. In the real world, even simple tasks end up being far more complex than anticipated and failure is costly. Seemingly simple automation tasks even hide complexity. For example, MSN replaced journalists collating stories with an algorithm and published and even preferred fake stories about mermaids and Bigfoot.
Experts Downplaying Problems
The scary part here is that when it comes to cutting-edge research in this area, where a system can get things wrong and cause a large amount of harm in the real world, many experts don’t seem to think it’s a problem. They aren’t acknowledging the threats and issues.
Even incredibly smart people don’t seem to grasp the impact of the issues. This is concerning because some of the very same people are building this technology.
Below are some statements from Yann LeCun, Chief AI Scientist at Meta. The first shows a lack of understanding of how misinformation and disinformation spread on social platforms, and the second makes a stunning false equivalency.
As far as his comments on generative art, I wrote an entire article on this subject. This statement is a false equivalency.
Now, I have a lot of respect for Mr. LeCun, and he’s certainly not the only one spouting these opinions publicly. But I’m using his case as an example to make a point. Experts developing this technology should take feedback from professionals in adjacent disciplines to strengthen their systems, not pretend they are only a few tweaks away from utopia. What has been happening is people take criticism, feedback, and misuse as attacks, instead of the healthy criticism necessary to improve systems. This was especially true with Meta’s Galactica, where genuine dangers were played off as insignificant mischief.
When generalizing to the real world, many of these issues are deeper and more systemic than anticipated. We aren’t merely a couple of tweaks away from fixing these issues.
Issue Handling Strategies
Current issue-handling strategies for these types of systems meant to generalize to the real world are sub-optimal. They typically fall into two categories, trapping conditions and adding more models.
Manually trapping conditions for all the things you don’t want a system to do is not a realistic or sustainable way of handling issues, especially with something as generic as a language model. Yet, many feel this is the way to go and we can see it in the way the engineers handled issues with ChatGPT as the situation evolved. But, you leave the door open to the human imagination and manipulation. The possibilities seem endless and you can only write so many rules.
Complexity is both the enemy of security and safety. This is something engineers should keep in mind as they look to engineer protections for their systems. On that note, another strategy thrown around is to train additional models to detect specific issues from the model. So, now we end up with an army of models with the purpose of keeping other models and systems in check, creating an opaque landscape where the next surprise is just around the corner.
More research is being done in this area, but today the gaps are still large enough to drive a truck through.
What Can You Do?
Always keep in mind that technologies like ChatGPT are experimental technologies, and you shouldn’t throw them into production systems. Unfortunately, people are going to do it anyway. Below are a couple of things to keep in mind before considering an experimental technology in a real-world system.
What’s the cost of failure in your use case? If it’s higher than insignificant, don’t implement the technology because it will fail.
Analyze known failures and trigger conditions for those failures. Try to recreate them. If, at first, it seems they are fixed, try different methods to trigger the same result. You could bump up against some condition trapping, and bypassing it could be trivial.
Perform extensive testing, including a series of tests specifically to get the product to fail. Be realistic about the results.
In short, you have to know the impact of your system failing and the various ways to trigger failure conditions in your system. If the problem space is narrow enough, you may focus the technology and trap specific conditions. If the cost of failure is insignificant, then you have some breathing room to experiment, but your mileage will always vary.
Conclusion
With this post, I just wanted to make a couple of observations on ChatGPT. It really is legitimately cool research and it’s moving state-of-the-art forward. Genies don’t fit back into bottles, so it’s important that we take steps now to plan protections to mitigate harm. Experts developing this technology need to take concerns and feedback from professionals in adjacent disciplines to ensure existing harms are reduced and safer systems are created as a result. The next few years are going to be interesting for sure.
The art world needs a revolution! A revolution fueled by randomness. Forget predictable brush strokes and perfect technique. Let’s just throw some words into the grinder, add the element of danger and see what happens. Why not? It’s Friday, it’s been a long week, and I needed a laugh, so I give you this gem.
Since the olden days, I’ve got a chuckle from doing something like catting /etc/shadow and piping it to your sound driver. It’s not like it makes beautiful music. It’s, well, kinda stupid. I wanted my own stupid thing. Well, it only took a couple of decades, the invention of cryptocurrency, and advancements in generative AI, but I finally have it. Welcome to the world of Crypto Wallet Pneumonic Generated Art. It’s cultured, refined, and, don’t forget, incredibly stupid. Yes, now you can risk it all to generate art from your wallet seed phrase. Behold! The beauty!
The image was generated with the following seed phrase.
This is all fun and games, but what if you did it with a real wallet? Hmm, let’s do that.
Here are 5 images generated from the 12-word seed phrase of a real Phantom wallet. How much SOL is in there??? Hmmm. You’ll have to recover the seed phrase and find out. At the time of this writing, the crypto market is crashing, so far less value then you’d hope.
Two of the Twelve words should be painfully obvious. Interestingly enough, from a few of the experiments I did, it seems to latch on to just one or two of the words that it’s most familiar with and discards most of the others on every subsequent run. It also seems to make some relations between words that explain how the model is constructing the image. I may run some further experiments using DALL-E as well and compare them to Stable Diffusion, but I’ll leave that for another day.
Time To Make Millions $$$
Now, for the business plan. What if NFTs were created based on pneumonic phrases from cryptocurrency wallets with balances in them? I mean, because why not? Since NFTs aren’t actually about art, this could be a way of superficially pumping their value. Will this be the next art revolution? Will people walk around galleries sipping wine and marveling at these masterpieces? Just asking questions 🤣
Try it yourself! Throw caution to the wind and join the artistic revolution.
Disclaimer: Not responsible for loss of funds due to complete stupidity!
