Home Creators Posts Import Register Favorites Logout

Content

This week's episode of Unpacked is now available!

Files

Big Walk's Big AI Controversy | Unpacked

This video is brought to you by The Snackbar at the End of the World, a bittersweet story of love, defiance, and… pasta? Take over a struggling restaurant to save your family in this hand-drawn, narrative adventure RPG set in magical 1960s Italy. – https://store.steampowered.com/app/5211590/The_Snackbar_at_the_End_of_the_World/ This week on Unpacked, we're getting nuanced over the big AI controversy surrounding Big Walk. Support us on Patreon: https://www.patreon.com/SecondWindGroup Second Wind Merch: https://shop.secondwindgroup.com/

Comments

shadowscribble

IMHO this is actually using AI responsibly. But since we have no culture of documentation it lead to the use being a surprise. Plus we need to acknowledge this knee-jerk refusal to accept these nuances Is Not Healthy. We've created an environment where devs don't want to disclose even the tamest use.

Mirko

You cannot use a "AI chatbot" ethically, for nothing, and that is by design. Your personal outcome doesn't reflect the actual costs, so this "10 lines of code" has quite a manipulative taste (since it serves as an mechanism to play down the impact). The story House House is telling, sounds a lot like "AI was our last hope to solve that issue". And that is just not true, it never is. To go and use _this_ kind of problem solving is a decision, not a necessity. That this decisions are very popular in software development is a pity, but really not all that surprising to me. The digital ecosystem often looks like an even faster and more exploititive version of the capitalistic system it lives in. Thanks for the nuanced comment, Nick and also for your consistency on that topic. I do struggle with the "it's shitty for art, but okay for code" stance tho. Apart from the costs of running autocomplete on steroids - which are the same destructive forces as when used for anything else -, it brings all the qualities into the process that it brings into other forms of human creation. The slop, the dumbening, the legacy, the excessiveness, the wastefulness, the psychotic effects etc. But it seems like code gets a pass, because all that it's measured against is the superficial functionality in that very moment.

Second Wind

I tend to think about something like Siri when talking about these things, and all the training on data it has gotten just from me asking it questions. It’s weird, I don’t know how I feel about it, and we interact with unethical things every day just to get through life. It’s all weird, and I think anyone claiming to have a definitive answer didn’t think about life enough lol.

Mirko

I was about to mention that these comments are written on devices, whose production and impact ranges from questionable to horrifying. But when it comes to whataboutisms, I think it is important to distinguish, who can take responsibility and to what extend. I cannot live a perfect life in an imperfect world. Neither can a company. But they do put stuff into the world on a big scale and their decision impact is magnitudes higher than that of an individual. So accountability and consequences need to matter significantly more. Not that hateful pushback is justified. But I do understand the anger. I am angry at any entity with power or reach that does not critically engage with problematic developments, when they could (and should). And in contrast, that's why I am very happy with Second Wind. ^^

majorenglishesquire

If a studio is at all interested in centering their human creators, they need to acknowledge that these problems existed before AI and were fixable before AI and if the fix to that problem was "turn off the shadows during sunset" then the omission should have been the answer. At this stage, where we are being beaten about the head with insistence that we capitulate and integrate AI into our lives even when it is of no assistance, I am not interested in giving oxygen to AI in any form. There is an active search for "morally acceptable" AI use cases. Little "oopsies" like this are what AI companies will grab on to and fly up the flagpole, establishing themselves as irreplaceable. As a target consumer, I'm simply not buying it. These problems existed before AI. Tracing back to the source of AI's build and development, the desperate, precious grab for "ethical" use cases is a non-starter for me. Theft was committed at a massive scale to build those structures. We will never be repaid for the theft of our labor that was used to build AI. None of us--not writers, artists, coders, photographers, voiceover artists, anyone. So. Tiny and undramatic as Big Walk's issue may seem, I think it's a tendency in developers that should certainly be snuffed by socmed controversy, unhappy buyers, public backlash, and hopefully then a pledge against AI going forward. It's easier to snuff a candle than a forest fire. I appreciate the thoughtfulness of this video (and I'm glad you have the stance you do, Nick). I am just disturbed in the extreme that it took so little time for people to forget how many of us were stolen from. A corporation with value in the trillions found my work worth training models on, but they wouldn't pay $3 for the ebook. The devaluation of people is not okay with me when so many of our lives are being upended by oligarchs.

The Maintenance Man

I'm a software engineer and an AI curmudgeon. Outrage over using a tool to solve a technical problem is a clear lack of understanding of how software is developed. Almost every application is made with an IDE built by someone else, using libraries built by someone else, on hardware built by someone else. The running joke before GenAI was that more than half of software development was googling questions, finding a post on Stackoverflow, and then pasting the answer into your own code. People have used non-generative code analyzers for ages to find behavioral bugs, security risks, and performance issues. The AI industry as a whole stands to destroy our environment and society and I hope we find a way to limit it. But clutching pearls that a coder didn't write every line of their application when they never did in the first place is just silly.

Tucker Labs

I think AI as a tool is fine, AI as an artist is not. If they tell us how they use it, the same way they tell us about the engine used and etc., then I really don't see the problem. Social media has just created so many attention hungry haters.

Tucker Labs

An indie developer doesn't have that much influence man. Transparency is good enough for me, I won't hate indie developers for using AI. It does suck that large companies are forcing AI down our throats but there are places where it is helpful and even life changing so it shouldn't be outright demonized as a tool based on what the worst people in the world use it for. I think we should have issues with the people creating the tool instead of the tool itself. But debate is healthy for society.

Davsau

This technology has serious social, economic and environmental costs. The companies that are foisting it on us all have committed a litany of crimes against all of us. And the tech billionaires profiting from it all seem to fall somewhere on a Venn diagram of Nazi, pedophile and Nazi pedophile. So I don’t think there’s really an ethical way to use this technology right now. Maybe in the future, after the bubble hopefully bursts and sanity prevails. But this is probably the best case use under the current circumstances. So I’m not mad, just disappointed, I guess. I still like Big Walk, but I hope they’re more transparent in the future.

Jason Youngberg

I do development for a nonprofit. We pay for Claude and I've had to use it on occasion. Last week I gave it three snippets of SQL code I was using to manually fix some bad data and told it to "combine these into one process". It did and worked after some tweaking. Next week I'll have to use it again. "Go to [website] and tell me which parts in the CSS files I need to change so the background in this one section is this specific color instead of a gradient". The CSS files are thousands of lines long and made by multiple people who are no longer connected to the organization. Both cases it's a tool to do something I could do myself, just faster. No generative. Not theft. Just a time saver.

Jason Youngberg

I do programming and feel the same way. Before we got Claude, my boss always said "google it" when I had a problem. (It didn't work most of the time because we used a specialized 3rd party system.) Now it's "ask Claude" and I get better answers faster. I hate using it, but it's getting expected to know about it so I use it for my job.

Ronny Cook

AI has a use case to enable and leverage productivity, rather than the naive approach of attempting to displace people entirely. For example, on Friday I submitted new code to our company's repository and the automated AI sanity check on the code failed because of several potential security holes. One of them I initially dismissed until I considered how variable interpolation was handled by this particular scripting language. AI is a tool. As with any tool, whether it is good or evil depends on the people using it. The same knife that slices carrots can be used to stab another person. A hammer can be used to build or destroy. The difference with AI is that it potentially displaces one of the key things that defines our humanity - the ability to think - and in doing so the wealthy people setting the agenda now have an additional lever to screw over workers. It's not a new story; computers have already displaced enormous amounts of intellectual work over the last half-century. The resource use of the tech is another issue. AI token pricing is currently artificially low, and so much of the current AI DC construction is likely redundant - pure waste. Again, not a new story; read up on the resource costs of Block Chain tech if you want an unpleasant surprise. The nature of capitalism is such that these issues will eventually solve themselves... at, unfortunately, enormous cost. All of which is to say... the devs in this case used the exact use-case for which AI is best suited. The people damning this may not be idiots, but they ARE letting their prejuduces rule their opinions. As supposedly impartial judges, they should be ashamed of themselves.

George Dorn

I look forward to the day that the bubble collapses and the frontier labs all crash and burn. There are plenty of good open-weight models these days, with hosting that's essentially a commodity. Unlike the frontier models, these have been developed and trained with limited resources and are far more efficient as a result. Several are run by companies that cannot pull in billions of VC money to subsidize tokens, so their costs really are the costs of running the models. I priced out what it would take to run GLM-5.2 a while back, and while it certainly isn't cheap, it's a cage in a colo facility, not a data center ten football fields in size with an on-site power plant because the local grid can't handle it. There are also versions of these models distilled to be especially good at one thing (e.g. Unity code) and can thus be squeezed into far smaller hardware requirements. This can be done without burning down the rain forests, it's VCs from SV that are pushing for it to be bigger and bigger and burn maximum power, so they can 'win' the AI bet that's going to crash the economy. So yeah, when the AI bubble finally pops, maybe we'll get back to nuanced discussion on the matter.

TehManticore

AI Issues What can "Used AI" entail? Seems like the discourse is finally getting to the point where we are differentiating what use means. Has anyone categorised the uses? - Generative (Create Something): Makes content (graphics, audio and code). The input tends to be text input sentences ("Make a video with cats in santa outfits", "Write some Python code to make radio button selection"), but it can also Transform existing content "Add explosions behind me in this video", "Make this code show three different options to the user and they select one"). - Problem Solving (Change Something): Asking AI to solve a problem, such as "Find why this code doesn't accept the enter key and show me the fix", "Test this program for security flaws", "Show me better furniture arrangements of my bedroom". - Teach/Summarise (Change Me): Provides explanations of general concepts and information. These aren't exclusive categories. For every use of AI, there'd be a spot on a Triangle Plot/Spectrum of how much the use matches these 3 variables. We'd each have a moral stance on each of the 3 uses but I don't think an LLM can be developed to do any of these tasks without the (stolen) works that went into it to create it. I don't think AI can possibly be made to do only one or two of these use types effectively without having been trained on all the (stolen) information they've used to date. Acceptable Use I've kinda always approved of AI for these things: - Non-commercial Comedy. I LOVE people being able to use AI to make skits, memes, and all the things that make life funny and enjoyable. - Teaching. There's the saying "Don't blame the student, blame the teacher". For the 100s of psychological variables involved, we often struggle to learn something new because the content isn't delivered in a way that personally connects or make sense to our brains. And teachers (professionals and unintentionals) don't have the time, resources or skills to mold their communication to every audience member. Youtube fills this need so well because people can often find a teacher that speaks right to their mind on the chosen topics, but finding those teachers can still be a very time-consuming process, and not every teacher is effective 100% of the time for 100% of their listeners. AI can be a way for people to find the right expression or conceptualisation in the time they need it. It shouldn't REPLACE teaching, but augment it. - Poor indy creators. As a wanna-be game maker, I know my strengths are in coding and not in art. If I want to get my game/s out there, I'm going to have to accept very low graphics quality (ASCII), use free assets, learn to art, or wait until I've saved enough money to pay artists (and somehow figure out my trust issues on whether they've used AI for the money). All of these options have their costs and benefits, but for an indy dev's first 1 or 2 games, I kinda don't mind the idea of poor startups using it until they can afford to pay someone to do it. - Non-profits. Whether you agree with me or not, I'm not sure we could get consistently applied regulation anyway. Regulation Even if we COULD agree on what is Acceptable Use, and whether the different kinds of use matter and how to honestly attribute that use, there seems to be little faith that AI could even be regulated around the world. I say this to name the problem rather than to add despair: The cat is out of the bag. The big corps aren't the only providers of these services, and "home grown" LLMs and open source software means everyone can potentially have their own AI in their home that has no limitations that companies may put on them (or have put on them by legislation). There MAY be some ways to regulate AI with truthful pricing and even taxation. They might even be able to log every product with a publicly accessible NFT. But I think the water has been muddied and we'll never get it back to pure H2O content (Human Squared Original? I'm sure there's a clever way to term that). And we're always worried about the "crutch scenario": We fear that regular use will push us down the slippery slope of accepting uses in the future that we find distasteful now, and we worry that our personal or collective skills will be diminished by dependence. If we can actually agree on what to limit with regulations, what if in the future we've become so desensitised that we (or future generations) will allow terrible things to happen? What if AI disempowers us to the point of disenfranchisement? We have a lot to hope for and a lot to dread. And I think some of that dread comes from feeling distrust of our lawmakers; that those in power aren't currently making laws in our best interests. If they seem corrupted and to be serving other interests, in what climate could we trust the powerful and rich AI CEOs, too? Accusations Comment sections are full of accusations of "Bot!" and "This is AI", with little evidence to back them up. It's easy to accuse, and harder and more bothersome to disprove. So what do we accept as proof of AI use? Some of the companies have said they'll incorporate signatures into their generative content that will make it so other software can tell it was AI. But that only matters if A: We trust ALL AI model devs around the world to do this signature-signing, and we trust the 'signature-checking' software, B: It can't be removed from the content after-the-fact, C: That this works for text (They say they can do this based on word frequency use...but when those frequencies feed back into training, how is this empirical and sustainable over time?), and D: That we get comfortable with showing everyone our precious code....

Willhart

I'm more OK with AI in art than in code. Bad art won't kill people while bad code may.

CthulhumonWriter

Yeah I've developed my nuance with AI, both in the personal responsibility for individuals and corporations. The simple fact is, it has far too much support from the wealthy, and too much momentum. You can't just put the toothpaste back in the tube. Currently my most significant hatred of AI is when a raw result is presented as a finished product. There needs to be human involvement, at least to edit the result, correct nuance, inconsistencies and outright contradictions in these things, otherwise we only foster and encourage the toxic apathy of the mass audience that just accepts whatever it spews out and says "good enough"

csepp

Massively agreed. If there are people trying to organize at a higher level to stop the arms race, we should be *cheering* them, not complaining. Or for an alternative metaphor: this is like if we complained about anti-doping in sports, because "everyone else is doing it".