News   GLOBAL  |  Apr 02, 2020
 9.6K     0 
News   GLOBAL  |  Apr 01, 2020
 41K     0 
News   GLOBAL  |  Apr 01, 2020
 5.4K     0 

My team uses it all the time for our website and labels. We drop in the Google translate and then send the draft to our one of our francophone staffers to double check and recommend revisions. Normally we get maybe four words or six technical terms wrong, which isn’t bad when it’s hundreds of words.
We constantly get notes from our Quebec stores about how we did it wrong, and how it should have been. All conflicting with each other and our in house full time translator.
 
I supervise legal work across the country, so often have to review material that’s from Quebec which has been drafted in French and will be finalized and filed in French. Of course, I don’t speak high level French at the level required to review legal wroting. The big law firms now have LLM based translation tools to translate stuff for this purpose, and it’s miles ahead of where it was a year ago. I suspect translation as a job will be gone in a year or two, except for very specialized/technical translation services.

For stuff like email blasts to all employees, those are still translated by humans, but I doubt that will last. For email blasts to customers, the human translators will probably hang on a bit longer because of the higher importance of technical accuracy, cultural IQ, etc., which humans are still better at ensuring, but I’m not sure how much longer.

I think AI is overhyped, but it’s going to result in major and very fast change for certain industries and jobs. Of course, that’s not new, especially in the last few decades.
 
Google Translate is still 'old school' machine learning….
The entire Google apparatus is old school nowadays. The very fact that they did not lead the development and introduction of AI, and now they can’t even make a credible copy goes to show that their near monopoly in web search and office apps has created a big, complacent beast teetering towards obsolescence.


Paywall free: https://archive.is/RopFc
 
The entire Google apparatus is old school nowadays. The very fact that they did not lead the development and introduction of AI, and now they can’t even make a credible copy goes to show that their near monopoly in web search and office apps has created a big, complacent beast teetering towards obsolescence.


Paywall free: https://archive.is/RopFc
The 'transformers' of general purpose transformer (GPT) is Google research.

Google has a long history of failing to productize their R&D. They've had advanced AI models for quite some time, and they have just very recently been put in a position where they felt forced to ship products even if they are not fully baked.
 
Good. But now we must be fair, so any non-citizen who kills anyone with their car should also be deported. Otherwise we’re just picking on the Indian.

At the time of the crash he was a new driver and had previous violations. And should not have been driving. The trucking company should be blamed and punished as well. Other wise we are just deporting one bad driver and replacing him with another bad foreign driver.

These mickey mouse driving schools are a big part of the problem too. No wonder we got so many bad drivers on the road.

Driving schools selling shortcuts to insurance discounts and faster road tests, investigation reveals​

Schools submit false information to government saying drivers completed 40 hours of training​


 
At the time of the crash he was a new driver and had previous violations. And should not have been driving. The trucking company should be blamed and punished as well. Other wise we are just deporting one bad driver and replacing him with another bad foreign driver.

These mickey mouse driving schools are a big part of the problem too. No wonder we got so many bad drivers on the road.

Driving schools selling shortcuts to insurance discounts and faster road tests, investigation reveals​

Schools submit false information to government saying drivers completed 40 hours of training​


That's more an insurance fraud thing. They still have to pass the G2 test.
 
Rules vary by province but minimal training standards coupled with virtually no government oversight of both driving schools and commercial fleet operators (in Ontario anyway) result in what we get on the road these days. The Humboldt driver might not be a particular outlier in the CMV industry, but it can't be denied that he killed a lot of people and was convicted of a serious criminal offence as a non-citizen.

*****

I supervise legal work across the country, so often have to review material that’s from Quebec which has been drafted in French and will be finalized and filed in French. Of course, I don’t speak high level French at the level required to review legal wroting. The big law firms now have LLM based translation tools to translate stuff for this purpose, and it’s miles ahead of where it was a year ago. I suspect translation as a job will be gone in a year or two, except for very specialized/technical translation services.

For stuff like email blasts to all employees, those are still translated by humans, but I doubt that will last. For email blasts to customers, the human translators will probably hang on a bit longer because of the higher importance of technical accuracy, cultural IQ, etc., which humans are still better at ensuring, but I’m not sure how much longer.

I think AI is overhyped, but it’s going to result in major and very fast change for certain industries and jobs. Of course, that’s not new, especially in the last few decades.
Maybe it will help the instructions I receive with a lot of stuff I buy that seems to be translated into English by one of my dogs.
 
Not much to say beyond I think this is the wrong take. AI is as bad as it will ever be going forward, and we haven't even seen a lot of the products based on the current AI tools come to market yet. Even a fair amount of code can be generated by LLMs, and just reviewed by a developer. The productivity gains are non-trivial, and not limited to listicle generation.

It is also likely near as good as its ever going to get given the flood of absolute harbage that is being generated on the internet and will invariably be ingested into subsequent models.

In terms of generated code, that's laughable. I personally tried some of these tools and they're absolutely hilariously wrong much of the time, and severely hamper productivity if you're doing anything beyond some extremely basic demo, which can just be copy and pasted off stackoverflow or something.

And in terms of the future, it's always harder for humans to catch mistakes as a code-reviewer than writer, (even if it is an important process of development). Humans writing the code (or editorials, etc) and then having tools do analysis after the fact to help find issues will always yield a superior product. But that increase in quality also requires more human time, not less.
 
For example, let's say the choice is between $12B for a fleet of diesel submarines built overseas and a $200B nuclear submarine that includes the design of a new CANDU SMR as its powerplant all built in Canada (with say technical assistance from the French), I would pick the latter. NATO commitments only say we need to spend more. They don't say what we should spend on. We don't have pressing threats like most of Europe. So we should be using our spending on something close to an industrial slush fund.

Nuclear submarines are the most extreme example. But, for example, our space sector can produce satellites that compete with the Americans. Telesat wants to build a constellation that rivals Space X. We can't subsidize them commercially. But we can give them a very generous contract to provide the CAF bandwidth worldwide. Global Satcom (not from Musk) could be one of our contributions to NATO. There's other examples like this.

Do I expect the Liberals to be this smart? No. Probably not the CPC either. But this is ideally how you spend more on defence without it just being a fund to enrich American and European defence companies.
Our US allies might think that's playing it too close. I'm not sure our critics would agree that commercial satellites with a CAF use can be considered legitimate spending towards the NATO commitment.


Really, the NATO target for all members should be measured on OpEx, not CapEx spending. For example, the new submarines are said to cost about $60 billion, which sounds like a huge shot in the arm against Canada's 2023 total defence budget of $26.5 billion. But once those submarines are built, the $60 billion is gone from the books, and our NATO spend drops. Therefore the only true measure of NATO spend should be the ongoing expenditure towards keeping the CAF operational.

To be fair, I recall a story that Greece was applying the pension costs of retired military servicemembers towards their NATO spend; so I expect many NATO members are play fast and loose. Even in the UK, once you omit the expenses on their nuclear deterrent (missiles, SSBNs, etc.), their defence spend falls well below their 2% NATO commitment.
 
Last edited:
I'm not sure our critics would agree that commercial satellites with a CAF use can be considered legitimate spending towards the NATO commitment.

Why? We count that spending now. And so do our other allies. Paying for capacity is a defence expenditure even if not used. I

Really, the NATO target for all members should be measured on OpEx, not CapEx spending.

There's a specific sub requirement. The goal is to spend 2% on defence and 20% of the total defence spending on capital programs and R&D. Canada is unique in NATO in not achieving either target.

Also, focusing on opex exclusively will not give a real measure of capability. Countries would simply under equip and overpay. Which is to some extent what Canada is doing.

To be fair, I recall a story that Greece was applying the pension costs of retired military servicemembers towards their NATO spend; so I expect many NATO members are play fast and loose.

That's not fast and loose. Pension costs are the cost of having military personnel. They are supposed to be counted. We didn't do that before. This was one of the definitions changed by the Liberals in 2017.

Even in the UK, once you omit the expenses on their nuclear deterrent (missiles, SSBNs, etc.), their defence spend falls well below their 2% NATO commitment.

I don't know why you'd exclude the cost of a nuclear deterrent. It's not a small matter. The British and American and (to a lesser extent) French deterrents provide a nuclear security umbrella to the entire alliance and ensure we don't need to field our own deterrent or much larger and more complex militaries.
 
The British and American and (to a lesser extent) French deterrents provide a nuclear security umbrella to the entire alliance.
Not according to this guy....
Because it's not NATO that is nuclear armed but three nuclear weapons states whose resolve is in question when it comes to defending allied territory. It's a valid question whether the UK, France or US would trade nukes with Russia over Latvia or Finland.
😜
 
Last edited:
Not according to this guy....

😜

Absolutely. But some ambiguity is better than none. And for the Europeans, French doctrine is particularly scary to the Russians. And that ambiguity and deterrence is still more than the rest of us could provide conventionally.
 
It's not our best and brightest moving to the US because they "hate the politics here" and need to move to the dumpster fire of politics that is Florida.
 
It's not our best and brightest moving to the US because they "hate the politics here" and need to move to the dumpster fire of politics that is Florida.

The stats were national. The article just happening to focus a narrative on Florida. Nowhere in the article do they say the majority of people moving to the US, are moving to Florida.
 

Back
Top