mtg wrote:
The tone of the OP is the real problem. He goes from acknowledging that he “used to
be a doubter” to accusing those who currently doubt as being “uneducated morons”.
Trait typical of cults and pyramid schemes.
AI is not a gimmick. It's the real deal, and it's coming really fast
Non-Running
-
20
-
As a proponent of Neo-Luddism for over twenty years, I suspect the hour is too late to turn back from the inevitable economic and societal changes brought on by exponential advances in algorithmic machine learning.
The amount of money and resources currently being directed to advance and accelerate these new technologies are already beyond a scale where they can be properly understood and slowed for proper safety concerns. Not wanting to be a pessimist, but the speed of change coming in the next few years will be logarithmically too fast to wisely control. Unfortunately, the institutions we rely on to protect us from these external threats are inadequate because of the current systemic level of bureaucratic incompetence and mediocrity.40 -
mayesrandall wrote:
I predict that your causal model is also imperfect wrote:
you are trying to solve the wrong problem using the wrong methods based on a wrong model of the world derived from poor thinking and unfortunately all of your mistakes have failed to cancel out
I have taught AI and I think that mattioco is closer to the truth, especially with his reference to humans have causal models. However, that is only true for certain tasks. Some tasks require brute force where deep learning and LLMs have an advantage, but they are not reasoning or understanding cause and effect. Some tasks require pattern recognition such as predicting the next word in a sentence. There is a whole category of AI which was used from the 1950s to 2006 when deep learning arrived. It is also very important.
I think the best way to test your arguments is to use specific tasks for accomplishing a goal such as developing a self driving car or winning the NCAA basketball racket. So far your arguments are Al Sharpton kabuki dancing.
With technology, I work with S-curves which is separating the hype from the true state of engineering. When I get a cocky student or argumentative person I use the Socratic method and real world problems. For example, for self-driving cars explain how you think AI perceives and object and makes a decision or would solve the NCAA bracket. Let's see if you got it goin on or if you are bloviated gas bags.
Mayesrandall this is complete Mumbai jumbo. Ofc AI has to modeled wrt to the problem it is trying to solve. It’s pretty simple how AI would be able to self drive. You provide the state of the world, likely a big matrix of fixed size tensor data, and you experiment with different architectures that best learn to predict the next state of the world and act according to a policy, and you optimize wrt to a given loss function, maybe a combination of avoidance of cars, people, + distance from the center of the lane, etc. As this is hard to do without damages to users cars, this is done in simulation, and takes place maybe after an extensive pre-training on collected human driver data. And no, predicting the NCAA tournament is not something that should be easy or expected for any predictive model. The space of possible tournament brackets is way too large and the model only gets one chance to make a prediction. Humans cannot predict all of the games of the tournament so I don’t know why AI would have these oracular powers. Since you questioned the LLMs reasoning I thought it would be informative for you to know how one might find use of it. So I know that the number of combinations of first round NCAA tournament outcomes is 2^32 (obviously some outcomes have much greater probability than others given team strengths, so the number of likely outcomes is much smaller). I don’t want to work out all of the combinations so I am going to have GPT4 do it for me.
01 -
What is the total number of ncaa basketball bracket combinations assuming we only care about games from the round of 64 onwards. In your computation, please explain each step of the process.
gpt4:
To calculate the total number of NCAA basketball bracket combinations from the round of 64 onwards, we need to consider the structure of the tournament. The NCAA basketball tournament, also known as March Madness, starts with 64 teams. The tournament is a single-elimination format, which means in each round, half of the teams are eliminated until a champion is crowned. Here's how the tournament breaks down:
1. **Round of 64**: 32 games are played, and 32 winners advance.
2. **Round of 32**: 16 games are played, and 16 winners advance.
3. **Sweet 16**: 8 games are played, and 8 winners advance.
4. **Elite Eight**: 4 games are played, and 4 winners advance.
5. **Final Four**: 2 games are played, and 2 winners advance.
6. **Championship Game**: 1 game is played to determine the champion.
For each game, there are two possible outcomes: either team A wins or team B wins. Therefore, for each game, there are 2 possibilities.
Let's calculate the total number of combinations:
- **Round of 64**: \(2^{32}\) combinations, because there are 32 games.
- **Round of 32**: \(2^{16}\) combinations, because there are 16 games.
- **Sweet 16**: \(2^8\) combinations, because there are 8 games.
- **Elite Eight**: \(2^4\) combinations, because there are 4 games.
- **Final Four**: \(2^2\) combinations, because there are 2 games.
- **Championship Game**: \(2^1\) combinations, because there is 1 game.
To find the total number of bracket combinations, we multiply the number of combinations for each round together, as each round's outcomes are independent of the others. Let's calculate the total.# Calculating total number of bracket combinations
total_combinations = 2**32 * 2**16 * 2**8 * 2**4 * 2**2 * 2**1
total_combinationsThe total number of NCAA basketball bracket combinations, considering only the games from the round of 64 onwards, is 9,223,372,036,854,775,808. This immense number reflects the complexity and unpredictability of predicting the entire bracket accurately.
obviously this is a pretty basic example, but it got the job done and was able to explain it fairly well.
21 -
Mumbai jumbo is one of my favorite soups.
For self driving cars--matrix, hmm--you get F. You need computer vision, NLP, and rules based AI to complete the tasks.
For the bracket you have potential---The space of possible tournament brackets is way too large and the model only gets one chance to make a prediction. This is very important. That is why Warren Buffet was willing to give a $1 million to someone who got the bracket correct.
The limitation is the transfer problem. Each game is a separate problem. Also, the day before the tournament you only have statistics for the first round and no data on the future rounds to train the data. More importantly, Bufffet, like myself is also interested in fundamentals--are players injured, healthy, match-up problems, etc. This is not in the training data.
02 -
TMADDDHASFNE wrote:
Not wanting to be a pessimist, but the speed of change coming in the next few years will be logarithmically too fast to wisely control.
Like watching a rich azzhole totaling his new Ferrari
i agree with you
11 -
People said the same thing about genetic engineering which originated in 1973. People are still saying boo, trying to scare others. Name one problem we have had. People get injured by falling in the shower and break bones and go to the hospital and die from infections if they are old and have weak immune systems, not the fall. Taking a shower is more dangerous.
I'm curious what may happen if AI goes through another growth spurt. What's the worst case scenario?
00 -
mayesrandall wrote:
People said the same thing about genetic engineering which originated in 1973. People are still saying boo, trying to scare others. Name one problem we have had. People get injured by falling in the shower and break bones and go to the hospital and die from infections if they are old and have weak immune systems, not the fall. Taking a shower is more dangerous.
I'm curious what may happen if AI goes through another growth spurt. What's the worst case scenario?
1 Now they're doubting Electric Medicine as well since it was crazy hundreds of years ago and has come back recently and heavily uses AI.
2 If you use canned AI you get what you pay for.
3. To have a custom AI that works the way you want you to build your own open source and load it up with your data sets to learn, test, then synthesize data sets, boost data sets, ad infinitum every week forever. It's an unending process. No doubt this is a barrier to entry. The average person can't build their own AI nor run 2:10.
00 -
still bothered wrote:
There seem to be two types of people in the "AI doubter" camp:
1. Educated people who don't really understand how AI works or aren't aware of how quickly it is advancing.
2. Uneducated morons who aren't able to reason cause and effect properly or have any sort of grasp at the scale of the changes happening.
I used to be a doubter, but not anymore.
AI is progressing faster than even the evangelists predicted. With OpenAI's latest release (Sora), you can create any kind of movie or video with a simple text prompt. Just whatever absurd premise you can dream up, it generates a video in any style you like. Or just normal things, like drone shots of a city or coastline.
AI basically killed videographers, drone companies, animators, video editors, photographers, graphic designers and a bunch of related careers overnight. There's no future in those careers; they have about 6-12 months left of viability.
AI is getting very close to being able to solve abstract problems without using the material is was trained on. This is the key point. Soon it will longer relies on data inputs from humans, it will be able to reason on its own.
Lawyers, healthcare technicians, data analysts, accountants, finance workers, copywriters, people working in marketing are next. Better start preparing.
If you disagree, you're not paying attention.
porn
20 -
You say "AI is not a gimmick. It's the real deal, and it's coming really fast"?
For 99% of us today AI is a questionable toy, sometimes accurate, a gimmick in most cases.
Only the 1% has access to specialized AI computers that cost a fortune for engineers to build.
00 -
As a software engineer I use GPT almost daily to just explain concepts or lazily debug stuff.
If I was starting an app or a SAAS or something, I would use it to create the skeleton or even basic functionality.
It's very useful and it basically gets rid of the need of junior level programmers.
00 -
Chat GPT is like a human being, it knows certain things well, and other things not so well.
If you want an AI system to handle a specific area of expertise it may cost you many millions of dollars to build.
00 -
Seven Trillion Dollars isn’t Chump Change.
00 -
The question is where does Altman want to spend the money--applications and killer apps or on research? Historically money has not been the problem, rather the AI winters were because of hype--overestimating our understanding of how the brain works.
Similar link, but no fee00 -
news media needs heros and villain.
with open source a i there will be 1,000,000 of heros and 1,000,000 of villains of a i.
both human and machine.
it's wild.
like nothing we've seen before.
10 -
In the past World Athletics employed internet gig contractors to edit video packages and image galleries. Now AI is being used. Saves money.
10
What People Are Talking About On LetsRun