More facial-recognition bans, new creeper tool links girlfriends to past porno, Microsoft's AI school, and more Let's get right to it: here's your latest roundup of recent machine-learning related news beyond what we've already reported. Cheap human labor is remote controlling Kiwibots: Food delivery machines known as Kiwibots may look dinky and sweet, trundling slowly across the campus of the University of California, Berkeley, to …
I imagine a campus full of bored & drunk fratboys going around & abducting the kiwibots.
The engineers among them may modify the bots to include functionality never envisioned by the manufacturer.
The programmers might update the software to include functions only dreamt up in the manufacturers' nightmares.
The jocks just strapping a huge dildo/vibrator to the top.
Then re-releasing the bots back on the campus to enjoy the chaos caused by their modifications.
After finding out that the "pilots" are making so little I'm also envisioning a protest over the practice by folks strapping a sign to a toilet plunger that reads "Don't use me! My driver makes less than minimum wage!" to the plunger stick, then suctioning the plunger to the top of the bot so it advertises the situation to one & all.
I just had a flash: folks sneaking up on the bots, flipping them over, knocking them over, or otherwise making it impossible for the bot to go anywhere & then stealing the food.
Do the bot drivers have some way of signaling local folks for help?
"Help! Please put me right side up!"
“Leaders in the public sector are often faced with unique challenges when considering how to apply AI to improve the speed and quality of the government services they offer their citizens,” said Mitra Azizirad, corporate vice president for Microsoft AI marketing.
Here's a challenge... how about public sector leaders just don't bother with something that usually doesn't work properly (if at all), stop cutting corners in the name of topping up their pension funds, and employ real people. AI is proven time again to be universally shit outside it's specially crafted training set. Get off the fucking bandwagon. It's not going anywhere.
AI answers questions to which it finds answers in AI database. I guess that is what DeepMind does - it finds answers, previously contrasting certain actions with certain structured descriptions (in the form of text).
AI database consists of synonymous clusters, that is, groups of synonymous patterns derived from paragraphs of texts. For example there is a paragraph:
-- Alice and Bob exercise a lot. Their goal is to be strong. --
In paragraph 16 patterns: 3 nouns + 1 pronoun, 2 verbs and 2 adjectives.
- Alice exercises a lot
- Bob exercises a lot
- Their exercises a lot
- Alice exercises strong
- Bob exercises strong
- Their exercises strong
- Alice is to be a lot
- Bob is to be a lot
- Their is to be a lot
- Alice is to be strong
- Bob is to be strong
- Their is to be strong
- Goal is to be strong
- Goal is to be a lot
- Goal exercises strong
- Goal exercises a lot
These 16 patterns are combined into synonymous clusters. For example one of them has 4 patterns:
- Alice exercises a lot
- Alice exercises strong
- Alice is to be a lot
- Alice is to be strong
This cluster makes sense and can be used, in response to a search query. The other clusters remain unclaimed, i.e. random groups of patterns that are not required stay unclaimed (or what I call lexical noise are not used).
I guess that DeepMind finds the best synonymous clusters in their context and subtexts that are in demand for a search query; where
- context is a specific game situation,
- and the subtext is the history of moves in similar situations.
This is my patented AI database technology.
Do you see? What can happen if you, as OpenAI, do not see the words that make up the patterns and ignore their parts of speech?
For example, the words "exercise" and "strong" can be nouns, and then in the paragraph
-- Alice and Bob exercise a lot. Their goal is to be strong. --
there will be 5 nouns, one verb and f, a total of 5 patterns... And just one pattern for Alice:
- Alice is to be a lot.
That pattern makes no sense, right?
Parts of speech for words are determined by structured dictionary definitions. For example, the word "goal" has a definition:
noun
1.
activity demanding physical effort, carried out to sustain or improve health and fitness.
"exercise improves your heart and lung power"
synonyms: physical activity, movement, exertion, effort, work; More
2.
a process or activity carried out for a specific purpose, especially one concerned with a specified area or skill.
"an exercise in public relations"
once
1.
use or apply (a faculty, right, or process).
"control is exercised by the Board"
synonyms: use, employ, make use of, utilize, avail oneself of, put to use; More
2.
engage in physical activity to sustain or improve health and fitness; take exercise.
"she still practiced every day"
synonyms: work out, do exercises, keep fit, train, drill, engage in physical activity; informalpump iron
"she still practiced every day"
This structured definition is compared with the patterns obtained from the paragraph and it turns out that it is a verb, not a noun. This is my patented technology too.
The remote-control Food Bots reminded me of a conference some time ago, at which several folks, including some AI gurus, were asked too predict the world in ten years (this was over 20 years ago, so...)
One of the AI gurus had a double prediction: that "ten years hence" (i.e.ago) domestic androids controlled by operators in third-world countries would be replaced by fully autonomous ones.
Living in VietNam with suppliers in Laos and Kampuchea, I often cross the border. The cameras are clearly marked Logitech - all the same model.
VietNam, curiously, doesn't use cameras at Immigration points.
Each time I have to cross the border I make changes in my facial appearance. The equipment gets confused when hair length covers the ears; when facial hair changes; it also identify different teeth (I use dentures).
The all time winner is when the lips are in a smiling configuration, whatever the cause - natural or artificial (denture) - and invariably produces a No Smile command from the Immigration officil..
One operator told me that the picture database is in the USA and that late at night, when airports are open, the speed of response is slowed (remember there is a 12-hour time differential between Indochina and the Eastern Seaboard.
The most impressive is the Chinese system which is fast and is available on desktop screens and portable radio / video terminals that are carried by police of all types and other officials.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
