Is there an easy way to fix AI?

Why fix it when it works flawlessly?

That's ideal. Nobody gets in trouble.

01100010 01100001 01110011 01100101 01100100 00100000 01100001 01101110 01100100 00100000 01110010 01100101 01100100 01110000 01101001 01101100 01101100 01100101 01100100

Wtf i Love self-autonomus cars now

most of these ""AI IS RACIST"" shit stuffs are not about the algorithms but the datasets employed -- white people are often overrepresented in these. but then there are purely physical sensing problems; higher albedo (lighter skin) makes for a stronger retroreflection from headlights, and overall typical better contrast against surroundings. black people are simply harder to detect.

>black people are simply harder to detect.
so people are easier to detect

Why fix it?

Fix?

Attached: 1555589751195m.jpg (779x1024, 146K)

Sounds like it's working properly to me.

Are gorrilas overrepresented as well or do they just happen to look like gorillas?

Attached: 1555436349538.jpg (1024x634, 93K)

Surely how well you see someone in the dark is a social construct. We are all equal!

based AI

KEK

Shes literally making a gorilla face

Call it racist

Youll get your a.i. gf one day

Based Elon

Excellent

Attached: ComfyCow.gif (273x253, 1.99M)

(((Levin)))

>Majority of contestants were white
>Most winners were white or asian
Jesus fucking Christ

STOP THE FUCKING PRESSES

OH PRESSES THAT IS

Attached: 1553468211934.jpg (1024x742, 101K)

Quads of quality

Lidar doesn't have this problem. It's only an issue for self driving systems that use visible and near visible light cameras. Guess which one Elon Musk announced last week that Tesla will use? Being born in South Africa must have had an influence on him.

Attached: Check em.png (589x589, 418K)

or underrepresented, as in not enough training examples to tell them apart. though yeah, they kind of do resemble each other which obviously makes it harder.

probably more likely to hit dark colored cars too

Sheeeeit why do unbiased reality got to be so damn racist

As a computer scientist, I hate these articles referring to any new software as "AI." It's just regular software, there's no learning of self-improvement built into the software itself.

Attached: BK.png (125x125, 2K)

try not to look like pavement?

same reason i tend to hit people at night, dont wear dark colors and dart in front of a 8,800GVW vehicle traveling at 35mph.

Based cars

>self-driving cars may be more likely to hit you if have dark sking
STOP TRYING TO COMMIT HIGHWAY ROBBERY ON SELF-DRIVING CARS

Attached: 57.jpg (1585x1527, 140K)

Yes there is. Gas jews. Problem solved.

These robots are looking based to me.

There'll always be a large scale aggregate learning cycle as new versions come out, even if it's not deemed safe to allow an individual vehicle to adapt too much. But I mean at this point even EFI is doing a crude form of learning as it adjusts fuel trims, ignition timing, etc.

AI keeps doing this.

I've been wondering about lidar. Wouldn't there be an issue with interference if every car had it?

It's fucking LASER BEAMS

Oh wow

Attached: 1554926208175.jpg (500x500, 58K)

Attached: bcdfxhdkqhi01.png (2400x2000, 813K)

Attached: 1553398795798.jpg (425x286, 34K)

Based and checked

Attached: 1eb.png (1169x496, 322K)

F

Attached: 1538914096095.png (1000x818, 720K)

Yeah and my radar detector goes off all the fucking time from grocery store automatic door sensors. What's your point

It's sad, out own AI creations are trying to solve the problems we won't

RADAR DETKTORS ARE HIGHLY ILLEGAL

Attached: 1553531325608.jpg (337x408, 83K)

Duh. Object detection involves the measurement of light around the object. Pale to light skin has more of a light reflective quality then does dark skin. Simple fucking science. Make 'em have a big ole toothy smile when crossing the road or some shit. Figure it out.