Gaby A.I. Class Blog!!

[blog-for-ai-class]

I think it would help me feel like I am making progress to work on a blog as part of this group. It would also let me give myself permission to think about more random things that aren’t the focus of our projects, but might spark ideas.

I want to use this to make notes on things I read as well.

Anyway, here are my goals for the quarter:

- post here at least once a week on things interesting to me but maybe not to my classmates

- fix + improve home assistant

- make good progress on group project

- complete/make progress a project comparable to other home assistant project in scope

[blog-for-ai-class]

Artist Statement

I don’t think that this encompasses all that I am interested in, nor do I entirely understand what I mean, but here is my statement of research interests for AI Research Group: 

I’m interested in how people define ourselves through our mind, body, and environment. The premise of AI technology as a trajectory towards an accurate simulation of the human mind may be flawed, but our exploration the subject allows us to explore and conceptualize other experiences of reality shaped by other forms of cognition and modes of perceiving the world. (Whether that perception/cognition belongs to organic beings or machines.)

As a literal manifestation of the mind/body dualism engrained in Western/European culture, and an as a supposed mirror to humanity, AI offers an opportunity to explore how that dualism affects people’s conceptions of self and relationships with bodies in such a culture.

[blog-for-ai-class]

I turned a Furby into an Amazon Echo. I give you: Furlexa

By combining a Raspberry Pi Zero W (a tiny computer), Amazon's open-source Alexa Voice Service software, and a few other electrical components, I converted m...

[blog-for-ai-class]

Final project ideas

A home assistant with fictional needs tied to a fictional body that doesn’t match - inspired by the Cozmo app that has makes you tasks to fill  its various need meters. -I’m interested in things like phantom limbs, personal space, how self-conception ties to the physical form -AI is supposed to be an entity without a body but depicted as having emotions and relating to itself in the way of someone existing in a human body -The idea of a mind that clearly doesn’t belong to the body it inhabits

-I mainly want to do this idea because it is weird and funny - constantly refers (to pain?) in abstract body parts it doesn’t seem to have - “my left branchius, my dorsal portico, segmental roundel” - gives shrill and insistent instructions to shake it in order to “rebalance the humours in my internal bladders” - sometimes requires these kind of rituals to access assistant functions

Other ideas My “Justin” idea but as a peer-to-peer sharing economy app. If you own an Echo, you can rent out your services as a roommate/live in home-assistant operator.

Add the ability to customize the meanings of words and phrases of everything the speaker understands. Create a highly personal language to speak to it. Change the set phrases to use those words.

[blog-for-ai-class]

Synergy? (https://mobile.twitter.com/edzitron/status/951586020503375872)

[blog-for-ai-class]

(good point)

^This podcast episode touches on almost all of the same ideas surrounding home assistants on as we have in our discussions

[blog-for-ai-class]

Internet of Shit. 6,478 likes · 30 talking about this. I guess we better put a chip in it.

A funny set of social media accounts poking fun at the insane rush to put “smartness” in every possible consumer good

[blog-for-ai-class]

[blog-for-ai-class]

Thank you for integrating Amazon Alexa into your device. We are excited for your involvement with Amazon Alexa and look forward to your product going to market. This page provid...

From the above article: 

Do Not:

Do not use the terms “Personal Assistant” or “Virtual Assistant” to describe or in the same context as Alexa.

Do not refer to your relationship with Amazon or Amazon Alexa as a “partnership”.

Do not show children under the age of 13, or that appear to be under the age of 13 (“children”), using Alexa in any marketing materials. While you may show families or children in your marketing materials, children should never be shown interacting with Alexa, and Alexa services should never be portrayed as being directed to or intended for children.

Do not make any marketing of your device overtly attractive to children (e.g., the marketing should not encourage children to ask their parents to purchase your device).

[blog-for-ai-class]

Ready to assist you

[blog-for-ai-class]

AI Home Assistant Notes

- Latest hyped tech product, giant corporations in competition

- Google, Amazon, Apple, Microsoft

- Still in early “generations”

- Not many people are using them right now

Creator intentions - Intended to be part of an “ecosystem” of other networked appliances based in a house setting - Intended to encourage use of other products and services provided by the originating company and partners —- I think you can buy Amazon products directly thru the Echo —- Google Home obviously integrates with all the Google services, Calender, etc. —- I can imagined the program being biased towards recommending certain partnered products, businesses when asked for recomendations —- different smart home products will be compatible with different companys’ home assistants - Intended to impart a sense of increased productivity, ease busyness - Planned obsolescence, advancing AI tech, competition to add new features will mean new ones every few years

Potential Problems

- Can maybe become part of bot net - Is “always listening to you” - security problems: “can something your shitty teen says in the presence of the Echo come back to haunt him in the future?” - collected speech is used to train AI  and advertise to you better - what if you can’t speak, or don’t speak a supported language -  “sub-human thing that is subservient to you”

Design - appearance of home assistants tend towards smooth, blank, few seams - criticized as looking like “trash bins” or cans - small, but not comfortably portable, except for a few types - not to be picked up and thrown around - Mostly resemble things already present in home environments in shape and materials — Echo sort of looks like a mini air purifier or something —  Google Home and Apple Homepod resemble speakers —-  to mitigate expectations of actual capabilties of the AI? ——  “a speaker, but fancier” - users encouraged to use natural speech — We are told often how much effort went into the natural language processing algorithms — people will probably develop Alexa-specific lingo, etc. anyway - What they mostly do NOT look like: — a computer — a robot (varies) — a toy — a bunch of circuits and electronic components — a person or creature — nothing (disembodied presence) — something that is connected to a a huge network of computers and other things   — a thing that is actively paying attention/listening to you — how could the visual aspect of home assistants more ‘honestly’, if possible, represent its functions and capabilities - the Baidu Raven has an usual design. Hipster Fisher-Price, one is not sure exactly what it does just by looking

Why the personal assistant metaphor? - people who have personal assistants- important and powerful business people and professionals, people with disabilities - anthropomorphizing —creates sentimental attachment — altthough “personal assistant” has some professional distance built in - what other metaphors could there be — A little node of [corporation], right in your home —– Google Lobe ™ — Engine of Reminding - “personal” - belongs to you, serves just you Expected consumers - people who have a stable place to live, spare income — I can imagine someone getting evicted and sitting at the bus stop with their Homepod :(

In the future - Assistants using tone of voice and emotion as contextual cues — could also analyze your emotions over the day in different to give you suggestions, ask you to reconsider an action if it seems you are angry

- Amazon is developing home assistants that look more like robots

Other stuff “Using the voice to communicate with an outside entity makes for an intimate and innately human experience….speech is the quickest way to get a point across “ https://www.engadget.com/2016/10/05/personal-assistants-google-home-ai/

Assistant software enforcing certain modes of interaction: - I think of Google Home there is something called “manners mode”, where you have to say please, to train your kids - https://www.forbes.com/sites/kalevleetaru/2017/12/13/can-ai-assistants-help-curb-harassment/#1ea5e9e1843f — Google Home asks not to be referred to as sweetie, etc.

- I can imagine these things being incomprehensible, omnipotent objects of wonder to a kid

[blog-for-ai-class]

My cozmo project

The current state of my Cozmo program is a loop where Cozmo checks to see if you are watching them, says “no” in a Cozmo voice, turns their back to you, and moves away a random distance, then repeats.

My original idea was to make Cozmo be a sort of adventuring companion. They would check if you were following them as they led you in random robot directions.

Cozmo’s usual m.o. is hanging around near to you, setting off a cascade of caretaking instincts. I wanted to make a Cozmo that valued its privacy and didn’t want anything to do with you.

At the moment Cozmo’s program has them constantly checking if you are there, but ideally after a while when they think they are alone they would stop checking for you and do some incomprehensible robot ritual.

[blog-for-ai-class]

Obviously AI is on my mind recently but I saw this roach mech thing and it reminded me of it because 1. non-human pilot 2. exists more to do its own thing, whatever that turns out to be, rather than perform a specific task. Its not even doing cockroachy stuff because it doesn’t know where it is or what’s going on  3. dumb confused robot with incomplete access to/ability to process information about the world to exist in it effectively

of course it’s also kind of disturbing that the roach is basically trapped there

[blog-for-ai-class]

Mario Klingemann on Twitter

“Making Faces”

amplifying and playing with idiosyncrasies and artifacts of images produced with neural networks 

[blog-for-ai-class]

Inside SynTouch, the Mad Lab Giving Robots the Power to Feel

Want robots that don’t crush our skulls? Well, we better give them a sense of touch.

Not quite AI but still somehow thrilling:

‘Embedded in a dome shape are 19 super-sensitive electrodes, over which a skin of silicon is stretched. The space in between is then injected with saline. So when you press the finger against something, the electrodes measure the changing resistance through the saltwater. The robot can sense how much pressure it’s applying, and quantify the give of the material it’s applying pressure to.

To measure texture, the finger monitors for vibrations. “So if I slide over a textured surface like my suit jacket, you’ll see these high frequency oscillations that are essentially the texture that you get,” says SynTouch CEO and co-founder Gerald Loeb. “If I slide it over a different surface there would be a different pattern of vibrations which I could analyze to say the difference between, for instance, the wool of my jacket and the cotton of my shirt.”’

[blog-for-ai-class]

Mind vs. Machine reading

This doesn’t touch much on the main themes of the article, but a part I liked early on was about how the machine programs were good at steering conversations towards their strengths and knowledge bases and away from areas where they might become awkward, or out themselves as non-human. The author points out that humans often don’t know how to do this, or don’t think to do this when it could be helpful. Computers have better memories of their own past conversation experiences and easier access to those memories for the sake of analysis. Taking an online personality quiz the other day, certain questions made me realize I have all these weird blind spots about myself. Do I “prefer organization over adaptability”? I felt like I would need to reference a database of all my recent experiences in order to answer fairly. Maybe this means that computers have better self-knowledge, even if they don’t know what it is to have a self. It’s hard for people to isolate and analyze their experience without bias. Which reminds me that In the last 20 years people have gotten big into “the quantified self”, “a movement to incorporate technology into data acquisition on aspects of a person's daily life”, in another variation on the theme of “people imitating computers,” which I guess does relate to what the author is getting at in the article. 

[blog-for-ai-class]

Various physical phenomena proposed to explain why animals sometimes seems react to incoming seismic waves before humans do.

I was interested in the idea of “non-human-specific intelligence”, so I looked up useful things that animals (potentially) have evolved to do that would be beneficial to humans to be able to do but can’t. I guess sensing something and reacting instinctively isn’t necessarily intelligence? But it seems like part of it. If the animals are running away from the earthquake they must have an idea that it is coming from a direction and what constitutes a major enough tremor to justify leaving their territory. So they have a mental model of sorts? And teaching robots to do things like depth perception and understanding the dimensions of an object from limited visual information seem like things on the same level of complexity. 

Related to non-human intelligence, I remembered this sci-fi story I read many years ago-- I don’t remember specifics but I think it involves an AI sex bot leaving his human to go live in the desert. He deletes all his memories and personality and everything except for his most basic learning algorithms in order to develop intelligence and a sense of self from scratch without human intervention. I thought it was very cool.