As a point of comparison, the Llama 3.2 3B model is 6.5GB. The entirety of English wikipedia text is 19GB (as compressed with an algorithm from 1996, newer compression formats might do better).
Its not a perfect comparison and Llama does a lot more than English, but I would say 6.5GB of data can certainly contain a lot of knowledge.
From quizzing it a bit it has good knowledge but limited reasoning. For example it will tell you all about the life and death of Ho Chi Minh (and as far as I can verify factual and with more detail than what's in English Wikipedia), but when quizzed whether 2kg of feathers are heavier than 1kg of lead it will get it wrong.
Though I wouldn't treat it as a domain expert on anything. For example when I asked about the safety advantages of Rust over Python it oversold Rust a bit and claimed Python had issues it doesn't actually have
Well the feathers heavier than lead thing is definitely somewhere in training data.
Imo we should be testing reasoning for these models by presenting things or situations that neither the human or machine has seen or experienced.
Think; how often do humans have a truly new experience with no basis on past ones? Very rarely - even learning to ride a bike it could be presumed that it has a link to walking/running and movement in general.
Even human "creativity" (much ado about nothing) is creating drama in the AI space...but I find this a super interesting topic as essentially 99.9999% of all human "creativity" is just us rehashing and borrowing heavily from stuff we've seen or encountered in nature. What are elves, dwarves, etc than people with slightly unusual features. Even aliens we create are based on: humans/bipedal, squid/sea creature, dragon/reptile, etc. How often does human creativity really, _really_ come up with something novel? Almost never!
Edit: I think my overarching point is that we need to come up with better exercises to test these models, but it's almost impossible for us to do this because most of us are incapable of creating purely novel concepts and ideas. AGI perhaps isn't that far off given that humans have been the stochastic parrots all along.
My guess is it uses the same vocabulary size as llama 3.1 which is 128,000 different tokens (words) to support many languages. Parameter count is less of an indicator of fitness than previously thought.
That doesn't address the thing they're skeptical about, which is how much knowledge can be encoded in 3B parameters.
3B models are great for text manipulation, but I've found them to be pretty bad at having a broad understanding of pragmatics or any given subject. The larger models encode a lot more than just language in those 70B+ parameters.
Ok, but what we are probably debating is knowledge versus wisdom. Like, if I know 1+1 = 2, and I know the numbers 1 through 10, my knowledge is just 11, but my wisdom is infinite in the scope of integer addition. I can find any number, given enough time.
I'm pretty sure the AI guys are well aware of which types of models they want to produce. Models that can intake knowledge and intelligently manipulate it would mean general intelligence.
Models that can intake knowledge and only produce subsets of it's training data have a use but wouldn't be general intelligence.
Usually the problem is much simpler with small models: they have less factual information, period.
So they'll do great at manipulating text, like extraction and summarization... but they'll get factual questions wrong.
And to add to the concern above, the more coherent the smaller models are, the more likely they very competently tell you wrong information. Without the usual telltale degraded output of a smaller model it might be harder to pick out the inaccuracies.
Oh man, we really need this for telecom too. And even more important than their collusion is their political corruption, so the government protects them from competition and funnels endless billions in taxpayer money to them.
You know what actually works great for reducing the price of cell plans in Canada is a public option.
There's one province where people pay roughly half the national average for all there telecom needs - and that's Saskatchewan. SaskTel's mere existence keeps all the majors in line.
Manitobans should be absolutely up in arms over the privatization of Manitoba Telecom.
Canada has a ton of land and just not that many people. There's 3 + Videotron major telecom operators servicing a population the size of California spread out over a land area the size of the United States. The Capex and payback period for spinning up a new, competitive, national carrier is wild (there's a pile of bodies there proving my point) which makes this a perfect job for the government.
(Not all were economic failures: you can also be a successful and scrappy going concern and that gets you acquired by bigger pockets with a rubber stamp from government that thinks bigger firms are better for all)
Not by definition but by the nature of the incentives.
They face a lot less monetary risks and people are harder to fire. Not to mention the additional bureaucracy and administrative weight due to the additional laws and regulations that apply to public services and how being a part of the state means facing more administrators and bureaucrats.
> Canada Post is certainly not cheap for the type and quality of service that's provided.
Can you pay less to send something via UPS or FedEx? I'm pretty sure you can't. Remember Purolator is a Canada Post subsidiary so probably don't count them.
I'm not sure. But I also don't consider Canada Post to be offering comparable service to FedEx and UPS these days, too.
For example, based on my experience and the experiences of other people in Canada I've talked to, Canada Post seems to just deliver a pick-up notice slip, rather than the parcel itself.
It's particularly annoying when the recipient has been available to receive the parcel, but Canada Post doesn't even appear to make an attempt to deliver, although the slip is left behind.
The recipient then has to go down to a Canada Post outlet with the slip to actually receive the parcel.
On the other hand, FedEx, UPS, and other private sector (I, too, would exclude Purolator) couriers consistently manage to deliver parcels directly to me, right into my hands, even if it might occasionally take a second attempt. I haven't had to go to a UPS or FedEx depot in well over a decade.
In my opinion, delivering a parcel directly to the recipient is a different service than delivering the parcel to the same city and then requiring the recipient to complete the job.
>There's one province where people pay roughly half the national average for all there telecom needs - and that's Saskatchewan. SaskTel's mere existence keeps all the majors in line.
Source? Looking at sasktel's website and rogers' website, this claim does not pass the sniff test. Their "basic" plan is $60/month (BYOD) for 15GB of data, whereas rogers gives you 75GB for $65. Even if you don't use 15GB of data and that extra $5 brings you no benefit, the price difference is nowhere near high enough to justify the claim "people pay roughly half the national average".
My takeaway from that is that Sasktel/MTS were much cheaper for a comparable plan, but the average revenue per user is near-identical - so the effect is mostly just that Sasktel/MTS users got more data with their plans. (Which tracks to me, if you halved the revenue of a provincial-scale carrier, they'd be a huge money loser regardless of how many executive salaries got cut or efficiencies got discovered.)
Probably marketing/promos for new customers vs. actual realized, and I strongly suspect if you click through those Rogers plans you'll land on a much higher total cost.
I looked again after your response, the Rogers plans all state "for new customers only" which is probably the discrepancy. Every time I speak with my parents I'm shocked how much they're paying and I have to spend an hour renegotiating.
I live in Saskatchewan and pay ~50$ for 10GB of data/month. I'm not on a Sasktel plan, as I found a cheaper option through one of the privatized networks, however Sasktel has a new subsidiary: Lum Mobile, to compete with the cheap plans like what I'm using. I know someone who has some discount at Sasktel and pays ~50$/month for 75GB(!) of data.
When I was in toronto the guy at the hotel had a 306 (saskatchewan) phone number because it was cheaper.
Sasktel is a great provider. I have fiber to the home in my house, for instance.
Even the minor browsers, pretending to not be funded by ads at this point (while the VC capital is drying up) depend on one of the 3 browser engines, all of which are funded by ads.
Safari? Unless you're going to say that Apple gets the money for Safari through ads which, y'know, technically correct but disingenuous in this context, surely.
Google is paying Apple 20 billion per year in their search deal, which is 40 times more than what Mozilla takes.
Safari is funded ENTIRELY by Google's ads, also making a profit, and this is a fact. We can entertain a counterfactual, maybe Safari would still be funded without Google funding it with billions, but that's not the world we live in today.
Not the person you're replying to and neither do I fully agree with them, but brave haa had issues with their crypto (BAT) system. Nothing that appears purposely malicious but quite possibly misleading in some cases.
Yes, looking for fair trade labelled products is important, and I also believe it is important to look for the International Fairtrade Certification Mark ...
As with any such program, if it is implementing effectively then some companies will chafe at the requirements and try to develop their own fair trade program more to their liking.
I think that for the rank and file, not needing to use the device is fine. But if you are a manager/product owner/etc (especially over teams in charge of said devices) then YOU specifically should have to use google devices like 24/7 as part of your job.
I think it's unhealthy for those in charge not to use the products they sell, especially when talking about consumer goods.
I think those directors, VPs and PMs should be mandated to use their own products. Okay to keep iPhone, but use Android as well, at least for their corp works. The product quality of Workspace and Chrome is debatable but it's not even remotely comparable to those niche market products because everyone must use it in their daily works. Why not for Android and their hardware products? Even FB did droidfooding long time ago...
I worked at FB when GearVR came out, and I was the only person on the floor who had an Android device to test it with (so I'm not sure the whole dogfooding thing worked very well there either).
the argument isn't that all googlers should have to dogfood all google products. But people in charge of a product line should have to dogfood it.
If you aren't willing to dogfood the the product line you have ownership of and prefer to use other products, you shouldn't have ownership in it as you have no vested interest in really making it better.
I think if you work on a team for some product, it would be good to be a user of that product. Especially if you are at a high level driving product decisions.
I think if you don't work on a specific product it is good to have diversity. For example GMail developers should probably use a mix of OSes and web browsers. But most Android GMail app developers should probably be using Android GMail.
But of course you don't want to be blind to other players, copying good ideas is good. So you do want to regularly interact with the other options as well. But I don't think only using the third-party option is a good idea either. If you only use an Apple Watch you don't know how the Pixel Watch works and how it has different interaction patterns (which on its own is probably a good thing). Maybe the product leads of the Pixel watch should regularly alternate between their own product and other competitors.
Well, this wouldn’t be a very satisfying explanation, but these JSON objects are often represented as Python dictionaries and those can’t have top level arrays.
But then I saw that users can get it free by posting "about screenpipe 10 times on social media".
If you want ads, pay for proper ads! Don’t pay people to turn user content into sneaky ads.
I understand why people hate regular ads, but IMO affiliate promotion (when done without disclosure) and stuff like what you are doing is much worse.
reply