ertgbnm

8 ms·

61.
by ertgbnm 2y ago
No human being would exist in the scenario described in my opinion. The scenario pre-supposes that humans would need to contribute some kind of value in order to justify their existence. I personally doubt any human being would be able to g
62.
by ertgbnm 2y ago
You must have already answered that question if you choose to scan your brain. Otherwise, you wouldn't get your brain scanned in the first place.
63.
by ertgbnm 2y ago
There was a lot of emphasis put on insulation values, thermal mass, and temperature differential. Half of the dog houses were like 20 pounds and open to the air on all sides. We put a 150 lb garden on top of ours to act as a heat sink and t
64.
by ertgbnm 2y ago
Even engineering contests are like this. I remember back in college there was a sustainable dog house competition with a very clearly defined rubric. My partner and I min/maxed the hell out of it and won first prize and $1,000, while e
65.
by ertgbnm 2y ago
I don't see a contradiction. Self belief means you believe that you can accomplish what you set out to so are willing to bet on yourself and take risks. Feeling like an imposter is a motivational drive from the other end of the spectru
66.
by ertgbnm 2y ago
Pretty sure that it would require that P=NP if such an event happened. So if factorization was cracked, everything else would be too.
67.
by ertgbnm 2y ago
Except in the real world we are allowed to offload the computation to a computer and have more time to double check things. Nice solutions are necessary due to time and resource constraints that exist within an educational setting.
68.
by ertgbnm 2y ago
I'm sure humans will always enjoy chess and art regardless of how much better AI is at it. In the same way, there will probably always be mathematic hobbyist who study math for fun. But I seriously doubt that in the near future there w
69.
by ertgbnm 2y ago
The problem is that Olympians are kind of just by definition outliers. Is this competitor a cheater because he can sustain 20 watts higher than others for most of the race? Or is he just a physical outlier who made it to the Olympics becaus
70.
by ertgbnm 2y ago
Yes it's easier and not too different from Turing's argument. However, Turing made this proof in 1936, you know, before the concept of pixels exists.
71.
by ertgbnm 2y ago
Private equity has mastered this formula in the last decade. Throw as many unrelated businesses that you can buy for pennies on the dollar into a basket. Claim it is now diversified. Decrease expenses by allowing all of your good employees
72.
by ertgbnm 2y ago
Pretty sure we could call an AR-15 vegan too. I think that's why it's weird.
73.
by ertgbnm 2y ago
The vertical monopoly is the big key. A uni lab couldn't replicate that.
74.
by ertgbnm 2y ago
https://www.youtube.com/watch?v=X5F2X4tF9iM This is what introduced me to them. May be a bit outdated at this point.
75.
by ertgbnm 2y ago
Same reason we might choose to have sex instead of masturbate. Masturbating may be faster and easier. But sex is fun and you can do it with friends.
76.
by ertgbnm 2y ago
Archaeology is almost exclusively circumstantial evidence...
77.
by ertgbnm 2y ago
This may be the type of thing that LLMs are currently the worst at. I'm not surprised at all.
78.
by ertgbnm 2y ago
I suspect Phi-3 is not robust to normal human input like typos and strange grammar since it's only trained on filtered "high quality" tokens and synthetic data. Since it doesn't need to waste a ton of parameters learning
79.
by ertgbnm 2y ago
That's the human's job for now. A human reviewer might have trouble catching a mistake, but they are generally pretty good at discerning a report about a mistake is valid or not. For example, finding a bug in a codebase is hard. B
80.
by ertgbnm 2y ago
You are missing the point of the model in the first place. By having higher quality RLHF datasets, you get a higher quality final model. CriticGPT is not a product, but a tool to make GPT-4 and future models better.
81.
by ertgbnm 2y ago
They used two non-mutually exclusive techniques. Phi-3 is mostly a curriculum training breakthrough. By filtering training set for high quality tokens and training on synthetic data, they were able to achieve great results. Gemma-2 is a dis
82.
by ertgbnm 2y ago
Eh, it's cute and I seriously doubt he is using this when not at home since it is a single use device. But also I am shocked that there is a New Yorker that would pick up a hat from the street and put it on their head. My first thought
83.
by ertgbnm 2y ago
I give it 2 years before they have regressed to a subscription model.
84.
by ertgbnm 2y ago
Yes, the Cretaceous–Paleogene extinction is agreed to be 66 mya with some surprisingly tight error bars of about 50k years. But for the purpose of discussion 100 mya is close enough in terms of orders of magnitude so there's not a need
85.
by ertgbnm 2y ago
Prefix notation isn't unconventional and it has many benefits. The main one being if you want to add many many items together it's much easier IE (+ 1 2 3 4 5 6). Instead of doing an operation between two numbers you are apply a f
86.
by ertgbnm 2y ago
$30 / million tokens to $5 / million tokens since GPT-4 original release = 6X improvement 4000 token context to 128k token context = 32X improvement 5.4 second voice mode latency to 320 milliseconds = 16X improvement. I guess I g
87.
by ertgbnm 2y ago
Over a year they have provided an order of magnitude improvements on latency, context length, and cost, while meaningfully improving performance and adding several input and output modalities.
88.
by ertgbnm 2y ago
I think people would have a deeper appreciation for what a super market is because they could understand it since it's exactly the kind of thing they would imagine a utopia would have. Throughout most of history people have spent most
89.
by ertgbnm 2y ago
More than anything else, I think the modern American super market would blow the minds of anyone born before 1900 more than any other marvel that exists. You have blueberries for sale in January??? A variety box of tea from 7 different coun
90.
by ertgbnm 2y ago
I'm not even going to discuss whether or not things are actually slowing down since I disagree with that. But I do feel confident that an AI winter is not on the horizon solely due to the overhead of implementation that we currently ha

More ›