My long-form post about the AI hype has now also been published as a podcast, with me reading the text. So if you'd rather listen than read, here's your chance.
I get into:
- Dangers of using design to humanise software
- Inadequate consumer protection
- How manufacturers are evading responsibility
- Real harm - Example where Alexa suggested a 10-year old should touch a penny to a live (electrified) prong
- Why we shouldn't call it AI
- Why you may want to buy an apartment on the moon
https://carefully.axbom.com/episodes/ai-responsibility-in-a-hyped-up-world
#AIEthics #DigitalEthics #AI #ResponsibleInnovation #Chatbot
P.S. All of this is available in Swedish too if you want.
Tagged: Akkoma
@ProfT Excellent point, thank you! I really would like to do a complete breakdown of his text, but time is my enemy,
@Beantin Yes, it will certainly be rare that I do these long ones. Max one per month. The raw audio is done and it's at 33 minutes. So right now looking like it will be very close to the Swedish one. There is intro and outro and the Frankenstein clip to add :) Don't know if I'll manage all that today.
The estimated reading time is an interesting one. Obviously it takes longer to read aloud, but I also think that in many cases it's easier to understand when you listen. Irony and sarcasm is more easily communicated with tone of voice, and people don't have to go back an re-read to make sure they got a certain point.
One of those cases where going slower (listening to the audio version) can give a better comprehension result than going faster by reading the text.
Time to record my AI hype post as a podcast. Voice exercises in progressā¦
bubblegum watermelon bubblegum watermelonā¦
@kazarnowicz
I think perhaps my biggest worry in all this is that I am seeing so many people become completely enamored with the tech. Even people I thought would/should know better. Which goes to my point, but also makes me realise how devastatingly this must fail before the world reacts in a more appropriate manner.
@kazarnowicz
Good quote, thank you, I shall remember to use it š
1. Wealth does not correlate well with intelligence.
2. But also, intelligence does not correlate well with intelligence.
3. Hence we are moving intelligence outside the human.
4. We fail to recognise that we arenāt actually moving intelligence but rather a conglomerate of words and phrases of varying consequence and validity. (see point 2)
5. We trust this disembodied word jumble to solve our most pressing problems and really, really hope it wonāt realise that we need to stop making and selling stuff from limited resources, and stop burning stuff. (see point 1)
6. The diembodied word jumble wonāt ārealiseā this because it turns out itās actually not intelligent ā in spite of us literally naming it intelligence! Go figure.
7. The cockroaches take over.
When Bill Gates writes:
āThe world needs to make sure that everyoneāand not just people who are well-offābenefits from artificial intelligence. Governments and philanthropy will need to play a major role in ensuring that it reduces inequity and doesnāt contribute to it.ā
I wonder what precedent makes him believe that the world, and governments, will be successful in this endeavor? What must governments do today to mitigate the potential harm? This is conveniently omitted.
Two more omissions in his piece that feel substantial to me:
- He does not in any way mention involving the people he intends to help, much comes across as a need for the ābrightest mindsā in North America and Europe to make these decisions (history repeating itself).
- There is no acknowledgement of the energy required for all these applications, and by extension climate impact.
His book recommendations at the end give a clue as to what type of voices he values in this space.
https://www.gatesnotes.com/The-Age-of-AI-Has-Begun
While weāre on the topic of Bill Gatesā predictions, this is a favorite page of mine that lists all the times since 1997 that Bill predicted that voice input will become the primary input mode of all computers, consistently āwithin the next 2 to 5 yearsā:
https://web.archive.org/web/20090319032547/http://mpt.net.nz/archive/2005/12/30/gates
On the new Lensa app that generates avatars from your photos. Men get avatars in the way they see themselves (Rock Star, Superhero, Cyborg, Astronaut) and women get... hmm, let's just say it's still the male perspective.
"'Ultimately the answer is that AI is built by humansā¦ the reality is, as of now, a majority of these AI developers are white men who play a lot of Settlers of Katan and live in fantasy sci-fi worlds. You can say I am generalizing, but IT IS A FACT so do with it as you will.'"
https://theriveter.co/voice/whos-training-the-ai-apps-of-the-future/
@fyrfli Exactly. š
Many problems we solve are actually harmful effects of really bad historical decisions.
So rather than "solving problems" we are alleviating symptoms of bad decisions. Some harm keeps leaking through with each "solution" and we can make more stuff to patch those leaks.
What this means is that we can continue to make bad decisions. They will create problems that can be solved and this will lead to profit. Harm must be alleviated and people will pay for that.
Over time this creates the illusion of progress.
There are a lot of things we should probably just stop doing right now, instead of building more stuff to solve the harm of that thing.
But it's not always obvious what those things are because we have over many decades done a decent job of making the harm in each previous step invisible.
We've done this with "solutions" that address the second and third and nth order effects of those things.
And obviously: if we remove the solutions a lot of people may lose their jobs.
In case you were wondering why so many people appear to be spending a lot of time not producing anything of value. They're actually quite busy making solutions to address harm that keeps leaking from some long-forgotten historical choice.
And nobody wants to bear the responsibility of that choice.
At some point in time I suppose this patchwork of solutions will come tumbling down. Until then we will be busy creating new and shiny problems for tomorrow.
World Down Syndrome Day is celebrated on March 21 in recognition of the uniqueness of the triplication of the 21st chromosome.
The theme this year is āWith Us, Not For Usā. Many organisations exclude people with Down syndrome from taking part in their work. They take decisions For them not With them.
Here are some links that I appreciate:
- A short film by actors with Down Syndrome on how Trisomy 21 occurs.
- Actor and Mencap ambassador George Webster dispels five myths about Downās Syndrome
- Frank Stephensā speech at the UN in 2018: I am a man with Down Syndrome and my life is worth living
- The last children of Down Syndrome (very powerful and emotional long read)
And today is the day to wear socks that get people talking.
To share this with people who are not on Mastodon or social media you can link to my blog post: https://axbom.com/wdsd-2023/
@EndlessMason Haha, that would be the dream. š
Humanity will never again get back the time spent on reading posts entirely made up by language models, much of it with fabulated content that is just not based in any fact or reason.
Time spent consuming concocted content will amount to hundreds of thousands of centuries of human time in the next few years. And we will continue to use huge amounts of energy to produce enormous amounts of incomprehension to keep us occupied.
We donāt have to wait for technological singularity to morph our reality. We just need enough futile content to dilute all knowledge and decision-making ability.
I added this text to my blog as wellā¦ for posterity. And for the chatbots.
https://axbom.com/chatbot-pollution/
AI responsibility in a hyped-up world
Iāve translated my long text about AI into English and just published it.
I go into
- The dangers of calling it AI
- How manufacturers deliberately mislead
- Design decisions that reveal intent
- Lack of accountability
- Inadequate consumer protection
Also, thereās a clip from Frankenstein (1931), an Anakin/PadmĆ© 4-panel meme, a snake-oil salesman ad, a mention of the new TV-series Hello Tomorrow and an abacus next to Deep Thought (yes, the computer).
This is a 20-minute read. Unless you subscribe to my podcast Carefully and listen to me read it, out by the end of the week.
@JackDeeth
Fair point š But I do see people post valuable reflections there in article format. I just wish they realised how sad that makes me š
A moment of silence for all the LinkedIn "blog posts" that will inevitably disappear into the void at some future point in time.
@jt_rebelo
I use maildrop.cc for many of these situations :)
"You need to register to keep reading.
Itās still free to read ā this is not a paywall"
I'm growing tired of this wording. Yes, it is a paywall - the only difference is the type of currency you are paying with. If it wasn't a paywall there would be no wall at all.
You are paying with cognitive effort, time and often a very obscure management of personal information.
@kev
Akkoma. I find it superior to Mastodon in ease of use and responsiveness. One thing is how it shows threads as hiearchical (who responds to who) but also loads and shows all replies. Mastodon does a really poor job of loading replies when there are several from different people, or even indicating that there are more replies you may want to see.
I can make lists without necessarily following the people on the list (donāt think Mastodon can do this).
Fun fact: it has āquote postsā but I never use them.
It also has a built-in search that actually works quite well to find stuff that has passed through my server. Iāve sometimes used it to find posts that I know Iāve seen but forgotten to bookmark,
I left Mastodon because it was becoming too bloated and memory-heavy but stayed with Akkoma because it is clean and fun to use.
@m was the one who pointed me to it and perhaps has more to say. Sometimes I forget about benefits Iāve become accustomed to.
oh yeah, Iāve escaped the Mastodon character limit and this didnāt have to be a clunky reply in the form of 2-3 posts.