That's Not Very Good Is It!
An Francisco-based self-driving startup Cruise announced it has entered a deal with Microsoft to use its Azure platform!
I anticipate lots of crashes!
Google last week locked the work account of the co-lead of its AI ethics unit, Margaret Mitchell, as she investigated the controversial ousting of her colleague Timnit Gebru. Late last year, Gebru, a prominent AI ethics expert, was pushed out of Google, which led to a substantial public and internal backlash against her …
I think it depends how it crashes. If the assumption of the operator is that it will crash, one hopes they will be more alert to things not going as planned. Also if it crashes the entire computer the car should stop and display a nice QR code on the console or even on the windscreen for the driver to scan.
Did anyone click the link to the PDF mentioned, that
It reads that Shmargaret Shmitchell is a co-author. Stripping the both SH's in first and family name that look very much like Margaret Mitchell..... Apparently Margaret knows the ethics within her employer!
So much for the AI technology of Google :)
"That suggests biases cannot be eliminated in the model itself, and will remain part of it."
That's hardly surprising - if you train it on what people say, then it will inevitably incorporate their biases.
What they need to do is find unbiased people to train it with. Good luck with that.
Just like I don't their machines reading my corporate mail that my idiot boss hosts on f*cken google server.
As being indirectly involved with a company that was suing Microsoft, I well remember the lawyers coming round to the people that I was hired to do the work for, while they searched for evidence in their emails. If the same thing happened today, and the company was using office 365 or whatever, would Microsoft even allow them to access their emails? For one sure thing, I would not be trusting that they would not be reading them.
I read the bit about "Microsoft, as the gold standard in the trustworthy..." and did fall out of my chair. If you consider MS as the gold standard of trustworthiness, your standards are so low as to make professional limbo dancers break out electron microscopes to determine if there's any clearance at all. I wouldn't trust MS as far as I could comfortably squeaze them out my urethra...
Vampire? Essex?
Would that be the MP for Witham, Essex?
This post has been deleted by its author
Why assume bad intent ?.. The AI software identified rabbits as the main culprit in eating salad plants of hobby gardeners.
Maybe the world is biased due to the events taking place.
Your remark " have looked at the data first .." sounds all to familiar, you mean censoring it, and leave out data you don't like ?.
Naive indeed. I suspect the majority of the people storming the capitol were christian, but you don't see them described as "christian terrorists". Even if you were training your AI solely on Guardian articles, you'd likely wind up with the conclusion "muslim terrorists" were more common than "christian terrorists" - whether it's true or not, the religion of brown people is considered relevant when it comes to terrorism, so it gets reported. The religion of white people is not, so it isn't (edit: that's a 2020's perspective of course, the religion of white terrorists was more newsworthy in the UK in the 70s/80s/90s)
As you clearly didn't like my last comment, let's take a quick count shall we?https://en.wikipedia.org/wiki/Terrorism_in_the_United_States#2010–19 will do as a source, because I'm trying to make a point, not do proper research.
Since 2016, in the US:
* 27 terrorist attacks
* 9 islamists (including people with slightly foreign sounding names where motivation is unknown)
So if you were training your data on this, you'd presume that terrorism was more likely to be committed by non-muslims. Maybe the AI is aware of this, it's not clear from the article. But it's fairly clear from your comment that you aren't. Maybe you should look at the data?
Not in our garden they’re not.
Ingredients:-
A skinned and boned rabbit, cut into bite sized chunks.
90ml brandy.
100g stoned prunes.
2 sliced shallots.
50g lardons.
250ml medium cider.
250ml chicken stock.
Flour for thickening, about 1Ts.
Butter......
You see where I am going with this.
Beer icon because it’s nearly culinary.
Sorry about that @Martin.
I’ve been cooking this for decades but the base line can be found in a “Hairy Bikers” on BBC food web site.
I also like using a rabbit/chicken combo.
In both cases you will need double cream.
Additionally, caramelise some apple slices and either put them on the plate as a side or mix them into the sauce at the end, along with the prunes and cream.
This weekend I shall be doing a venison stew, as a bourguignon, because it looks like being chilly again.
Oh, BTW always sneak some cognac from the bottle while cooking. Am I an angel?
Am I an angel?
Keith Floyd - RIP. Now Cooking for Angels.
Here's his take on the dish
https://www.youtube.com/watch?v=aAIdr-_ozGE
Hopefully the above link is not geo-fenced to Blighty. It's from BBC Studios
Agreed, Ghettoising minorities into special interest groups and echo chambers really is not a solution. Maybe something called Inclusive In AI would be a better choice. Otherwise all the Indian and far eastern people will still be excluded.
Look at MumsNet. It claims to be a "parents" group but it's very name excludes fathers. Any father who does join is looked on with suspicion and the first hint of disagreement with a "mum" and they get banned.
Methinks the likes of an OpenAI's GPT-3 language-generating model, and all major Silicon Valley type Players, no matter whereabouts on the Earthly globe they be based, also of a similar DeepMinded persuasion and addictively dabbling in the meme and technology, have the same abiding accompanying difficulty to reconcile and overcome in order to avoid themselves being recognised as an almighty alien enemy within, to be corporately crushed and virtually vanquished from future vitally important, extremely effective fields, if the past and present day current status quo establishments are to remain practically intact and still virtually in overall command and almost universal control rather than progress delivering something significantly better and brand spanking new for executive novel administration.
That difficulty is ...... ensuring Natural Language Processing in machines does not uncontrollably morph into Neuro-Linguistic Programming of humans.
There are those who would prefer to tell you that such is a real opportunity which will not be missed rather than a difficulty to encounter which cannot be stopped, for both are perfectly true but with the one being thoroughly enjoyable and the other certainly not so.
What say you? A designedly odd and decidedly strange opportunity to embrace and engage with, or a crazy resource draining difficulty to battle and be defeated by?
*. Or did they do it later with AI and IT ‽
Stop making sense making sense! ...... Someone Else
Whenever politically incorrect and virtually inept worlds are drowning in so much media hosted nonsense, Someone Else, one is almost duty bound to ensure and guarantee catastrophic collapses are mandatory and sure to be Autonomously Automatically Anonymously Provided with Advanced Instructions for Following and Trail Blazing/Trialling and Path Finding‽ .
The petrifying difficulty which creates a seemingly impenetrable wall to simply hurdle over for leading humanity organisations to progress and process novel information with Advanced IntelAIgents is the crazy abiding notion that the future should be similar to and pay endless tributes and crippling interest to the past ...... which of course is/was, as we now know, specifically designed to inordinately enrich and outrageously advantage corrupt and perverse subsets of a relatively small, and invariably self-chosen deluded few. .... where hubris rules supreme in the thrall of arrogance in those mindsets.
Fortunately though, A.N.Others are not so paralysed and powerless and clueless and would be into freely sharing solutions with everybody who is anybody anywhere whenever it be clearly known they need and want IT with Novel Solutions magically delivered.
And you don't get many being able to do that nowadays.
This sort of job sounds cosmetic, the sort of post you generate to fill a PR void due to a lot of popular misunderstanding about what AI is and isn't. I think the problem Google is having is merely that the people they selected to be "AI Ethicists" are actually contributing to the confusion rather than doing something rational about it. From a quick scan of the paper it appears that the authors are concerned about the size, cost and power consumption of large computing arrays and how they impact disadvantaged people as well as the imperfect results the models generate with the training data that's used to work with them. This is, unfortuantely, a fact of life -- there are few occasions throughout history where individuals could lay claim to the same quality and quantity of resources that large organizations can. This doesn't invalidate the work that's done on a large scale, especially as much of the large scale work will be wasted effort. (Far better for Google to waste energy on AI research than have us all have a go like we see with cryptominers.)
The other issue that's not covered here is the relationship between employers and employees in a corporation, especailly in a state like California. The employees seem to think that because they've got "Ethics" in their job title that this gives them some kind of special status. It might, but only by the grace and favor of the employer. Essentially all employees are 'at will' -- they can be fired at a moment's notice for any reason, the only distinction being 'for cause' or not which dictates your eligivility for unemployment insurance. Companies set the rules and companies are legally allowed to enforce them. Its not a satisfactory situation but its the law in our 'Right to Work' states. The original problem was caused by someone directly challenging company policy. They will lose; its not a matter of whether they're right or wrong, they will lose. Attempts to sue employers for unfair dismissal, discrimination or what have you also tend to be futile because the employer will have arbitration woven into its contract with the employee (which is shorthand for 'you lose').
(To those tempted by Silicon Valley bear in mind that the H1 is now a crapshoot, you've got no rights as an employee and if you do sucessfully run the gauntlet and get a Green Card your reward is that you're 'owned' by the Internal Revenue Serivce. For life. (Effectively.))