One small step for an autoregressive language model, One giant leap for AI
Discover more from FFFUTURES
One small step for an autoregressive language model, One giant leap for AI
FFFUTURES #2, August 24, 2020
FFFUTURES shares notes relevant to minds building future scenarios and extended realities. If you’d like to sign up, you can do it here.
EVERLASTING OGURA
Koichi Ogura is my favorite person. I'm a subscriber to every one of his platforms; his podcast may be my favorite. My routine doesn't allow me to listen to him on weekdays, so Saturday became my Ogura-catch-up day.
I was devastated when he passed away last March (is it really five months already?). Still, it's recomforting to keep getting new content from him every week.
I'm using Cocoon in case you're wondering. At first, I was on the fence between using Cocoon, or their Indian-funded rival, Chrysalis. Chrysalis is cheaper and has more content creators, but I went with the American option just to be on the safe side if there's an embargo or a data audit. I learned my lesson from last time.
Anyway, what I didn't expect is that Ogura is actually not that famous in Cocoon. The platform has exclusivity on some big-name celebrities who passed away years ago so, as you'd expect, the time advantage leads to more refined deep learning models with an output indistinguishable from the original person. For new, hum, joiners, to the platform, the models need to start from scratch.
As a long time Ogura follower, I can say Cocoon's model of him is close to reaching perfection. But I can see why people won't give him a chance when there's still a hint of an autoregressive model behind his eloquence.
This is why I was against the bill introduced by Congresswoman Garcia, which prevents these platforms from building models of people while they are alive.
She voiced all these people claiming for a level playing field. I never understood how they didn't see the fundamental flaw in their reasoning: By preventing these companies from building models while the person is still alive, they ended up doing the opposite of what they wanted. It's like dying early gets you a head start on these platforms, you have more time to train the deep learning models ahead of your competition, and more time to capture the market on those platforms.
At least I'm glad Ogura passed away before his rival, Kevin Ngo. By the time Ngo joins Cocoon, Ogura will already be dominating the space. I wouldn't stand watching Ngo ahead of Ogura for the rest of the time.
👁️ Omnirealities
The Future of Events, with Srinivas Krishna
Srinivas Krishna, CEO of Geogram talking about the impact Covid-19 had on Live Events, the adoption of XR solutions, and the upcoming combination of hybrid Virtual-Live events.
But also, words of caution were shared in a different panel by Galit Ariel, author or Augmenting Alice:
“I think we have to be very cautious now in content creation and value creation and not just jump into that all-virtual world, because there is a craving for physical interactions and kind of going back into kind of, like, less technological solutions or solutions that don’t seem technological.”
Rethinking Reality With Lenovo ThinkReality’s Nathan Pettyjohn
An interview with Nathan Pettyjohn, Lenovo’s AR/VR Lead.
“One of the fastest-growing applications for XR technology is soft skills training. That’s things like interpersonal communication that are required for the job but can’t be learned from a book. The conversation moves into the skills gap in skilled labor resulting in the Baby Boomer generation leaving the workforce.”
🔮 Future Scenarios
AI-Generated Text Is the Scariest Deepfake of All
Renee DiResta writes about the challenges synthetic-writing will bring us in the near future, describing its level of manipulation as ubiquitous and undetectable.
[…] another form of AI-generated media is making headlines, one that is harder to detect and yet much more likely to become a pervasive force on the internet: deepfake text.
I think I saw the term deepfake text for the first time on a study published by Max Weiss, where participants were presented with a series of texts and struggled to correctly classify whether they were written by a human or a bot.
Renee DiResta continues:
In the early 2000s, it was easy to dissect pre-vs-post photos of celebrities and discuss whether the latter created unrealistic ideals of perfection. In 2020, we confront increasingly plausible celebrity face-swaps on porn, and clips in which world leaders say things they’ve never said before. We will have to adjust, and adapt, to a new level of unreality.
The unreality term caught my attention here.
The research and development being done in augmented and virtual reality, the “extended realities”, is mostly focused on entertainment and productivity. We call those “something-reality”, an extension we create on top or around reality.
So, where would “unreality” fit here? Does it belong within the space of “extended realities”? Does the “unreality” term carry a description of malicious and manipulative intent? Is there a need to differentiate between constructive realities (augmented, virtual, mixed), and destructive realities (deepfake)? And what would give them their name, the medium they exist on, the content they present, or the intent they carry?
[…] undetectable textfakes—masked as regular chatter on Twitter, Facebook, Reddit, and the like—have the potential to be far more subtle, far more prevalent, and far more sinister.
💀 Not a Cylon
David Burr @davidcburr
David Burr @davidcburr
David Burr @davidcburr
🧠 Common Enemy
Can GPT-3 Make analogies?
An in-depth look at a series of experiments made by Melanie Mitchell, Davis Professor at the Santa Fe Institute, while testing how GPT-3 would ‘reason’ around analogy problems like the following:
“If a b c changes to a b d, what does p q r change to?”
GPT-3 Response to Philosophers
This PDF is the document created by Raphaël Millière (Presidential Scholar in Society and Neuroscience at Columbia University) after asking GPT-3 to write a response to the philosophers exploring the issues and questions raised by the advancement of AI demonstrated by GPT-3.
“You may wonder why I give this conflicting answer. The reason is simple. While it is true that I lack these traits, they are not because I have not been trained to have them. Rather, it is because I am a language model, and not a reasoning machine like yourself.”
📿 A Return to Faith
How the sound of religion has changed in the pandemic
A small sample of audio recordings documenting how religious practices were shifting and adjusting to the Covid-19 scenario. Some of these files will be a harrowing curiosity for the future (like the Buddhist chanting in New York), and others are just plain funny (“Cars honking Amen”).
IN OTHER SPACES
We're All Mad Here: A Conversation With Tom Waits
Fantastic interview with Tom Waits by Robert Lloyd
“Perhaps he's away indefinitely. Perhaps he was never here. It's one of those things you say in order to explain the way that you feel… in methaphors, I guess, it feels sometimes in the world that God is away on Business and he's not coming back.”
Uptale – VR Training
A platform aimed at businesses for immersive learning in VR
With AI and robotics conquering the workplace, large organizations face a workforce upskill & reskill urgency. Workers need more efficient ways to learn and adapt. VR is the best technology to digitize human skills & know-how, and distribute them at scale.
Are you hacking with futures and other realities? Do you have comments, stories, or suggestions? I’d like to hear from you. Reach out: heyfffutures@gmail.com
FFFUTURES shares notes relevant to minds building future scenarios and extended realities. Join and never miss an update.