thelittleshort
Top Bloke
I feel like I’ve posted this previously but maybe not
BRN Discussion Ongoing
- Thread starter TechGirl
- Start date
HopalongPetrovski
I'm Spartacus!
Extract from article from NY Times about the near future re Chatbot etc....
"Before GPT-4 was released, OpenAI handed it over to an outside group to imagine and test dangerous uses of the chatbot.
The group found that the system was able to hire a human online to defeat a Captcha test. When the human asked if it was “a robot,” the system, unprompted by the testers, lied and said it was a person with a visual impairment.
Testers also showed that the system could be coaxed into suggesting how to buy illegal firearms online and into describing ways to make dangerous substances from household items. After changes by OpenAI, the system no longer does these things.
But it’s impossible to eliminate all potential misuses. As a system like this learns from data, it develops skills that its creators never expected.
It is hard to know how things might go wrong after millions of people start using it.
“Every time we make a new A.I. system, we are unable to fully characterize all its capabilities and all of its safety problems — and this problem is getting worse over time rather than better,” said Jack Clark, a founder and the head of policy of Anthropic, a San Francisco start-up building this same kind of technology."
www.nytimes.com
"Before GPT-4 was released, OpenAI handed it over to an outside group to imagine and test dangerous uses of the chatbot.
The group found that the system was able to hire a human online to defeat a Captcha test. When the human asked if it was “a robot,” the system, unprompted by the testers, lied and said it was a person with a visual impairment.
Testers also showed that the system could be coaxed into suggesting how to buy illegal firearms online and into describing ways to make dangerous substances from household items. After changes by OpenAI, the system no longer does these things.
But it’s impossible to eliminate all potential misuses. As a system like this learns from data, it develops skills that its creators never expected.
It is hard to know how things might go wrong after millions of people start using it.
“Every time we make a new A.I. system, we are unable to fully characterize all its capabilities and all of its safety problems — and this problem is getting worse over time rather than better,” said Jack Clark, a founder and the head of policy of Anthropic, a San Francisco start-up building this same kind of technology."
What’s the Future for A.I.? (Published 2023)
Where we’re heading tomorrow, next year and beyond.
thelittleshort
Top Bloke
This NVIDIA GPU Firmware engineer has hands on Akida experience from Carnegie Mellon University
Fact Finder
Top 20
AKIDA has been implanted successfully in young minds.
These young minds have infiltrated the inner sanctum of Count Von Neumann.
The Revolution has begun.
My opinion only DYOR
FF
AKIDA BALLISTA
thelittleshort
Top Bloke
This chap was a consultant to Sony India working on SNN for object detection at same time as he was using Akida
Fact Finder
Top 20
Ignitarium Technology Solutions is an ARM partner and lists Brainchip as an Ai partner:
Ignitarium
Ignitarium is a boutique product engineering design house that combines its deep expertise in semiconductor, AI and wireless-enabled embedded systems.
www.arm.com
My opinion only DYOR
FF
AKIDA BALLISTA
Tothemoon24
Top 20
.
Qualcomm Expands Connected Intelligent Edge Ecosystem Through… | Amit Mate
GMAC Intelligence is excited to partner with Qualcomm as they expand the Connected Intelligent Edge Ecosystem through groundbreaking IoT and Robotics products.…
White Horse
Regular
There is something fundamentally flawed with this reasoning. It's a DUD.
Last edited:
Boab
I wish I could paint like Vincent
It is the 1st of April (Fools Day)
Most would not realise that the NASDAQ is in a technical Bull Market.
Its risen around 21% since the Oct'22 lows.
Most would say how could that be when the world is falling apart??
The answer is that the market does not give a shiete what people think. It just does what it does.
Interestingly the NASDAQ started the bull run in OCT'22 with the same price action bullish set up as we now see with BRN.
The XJO completed the same price action yesterday as well.
The last 2 huge BRN bull runs saw the same bullish price action set up in the lead up to those runs.
I am not making predictions or offering opinions. I am merely noting down what price action reveals.
Last weeks low of 47.5 cents must hold for the set up to remain in play.
All may be just a coincidence so DYOR.
Its risen around 21% since the Oct'22 lows.
Most would say how could that be when the world is falling apart??
The answer is that the market does not give a shiete what people think. It just does what it does.
Interestingly the NASDAQ started the bull run in OCT'22 with the same price action bullish set up as we now see with BRN.
The XJO completed the same price action yesterday as well.
The last 2 huge BRN bull runs saw the same bullish price action set up in the lead up to those runs.
I am not making predictions or offering opinions. I am merely noting down what price action reveals.
Last weeks low of 47.5 cents must hold for the set up to remain in play.
All may be just a coincidence so DYOR.
Dhm
Regular
Great video, but typical that it shows IBM and Loihi 2. They forget Brainchip and our first to commercialisation.
BTW I saw microcombs but I don't understand them. Perfect cue for our resident brainiac @Diogenese to enter stage right. Or others too if they know
And Swinburne University of Technology is Australian.
https://www.swinburne.edu.au
White Horse
Regular
Ya got me.I am floored by the flaw in your floor
Busy watching the footy.
Deadpool
Did someone say KFC
FrederikSchack
Regular
Hi Chippers,
I was just speculating a bit.
Akida 1.0 could run about a million transistors (depending on configuration) at a 32 nm process node, at milliwatts of power consumption. If we scaled down to the lowest process node that we have today, which is 2 nm (Intel just delivered some (probably laboratory versions) to the US army), then let's say we could pack 200 million neurons, still running sub 1 watt. Connect 5 of these and you would have a billion neurons running sub 5 watts.
Given that with some further development of technology that is already in existence, we may be able to reach a billion neurons working at GHz speed and sub 5 watts of power.
Considering that AI already is starting to help us develop AI, through for example Microsoft Copilot that assist programmers and Synopsys' AI assisted chip design. This is already creating a feedback loop, that will accelerate development of AI software and hardware. Given that it's probably technically possible to create a billion neuron network, running sub 5 watts at GHz speeds and the feedback loop. The probability of this happening in a few years seems high?
This system with a billion neurons will be running billions of Hz (GHz), compared to the human brain of a bit less than 100 billion neurons that runs roughly 50 Hz. So, it's likely that given the right model, it would be able to reach advanced conclusions that humans aren't capable of reaching. Not saying that it will be a substitute for humans, more that it could be vastly superior in reaching some conclusions?
I was just speculating a bit.
Akida 1.0 could run about a million transistors (depending on configuration) at a 32 nm process node, at milliwatts of power consumption. If we scaled down to the lowest process node that we have today, which is 2 nm (Intel just delivered some (probably laboratory versions) to the US army), then let's say we could pack 200 million neurons, still running sub 1 watt. Connect 5 of these and you would have a billion neurons running sub 5 watts.
Given that with some further development of technology that is already in existence, we may be able to reach a billion neurons working at GHz speed and sub 5 watts of power.
Considering that AI already is starting to help us develop AI, through for example Microsoft Copilot that assist programmers and Synopsys' AI assisted chip design. This is already creating a feedback loop, that will accelerate development of AI software and hardware. Given that it's probably technically possible to create a billion neuron network, running sub 5 watts at GHz speeds and the feedback loop. The probability of this happening in a few years seems high?
This system with a billion neurons will be running billions of Hz (GHz), compared to the human brain of a bit less than 100 billion neurons that runs roughly 50 Hz. So, it's likely that given the right model, it would be able to reach advanced conclusions that humans aren't capable of reaching. Not saying that it will be a substitute for humans, more that it could be vastly superior in reaching some conclusions?
IndepthDiver
Regular
Hi Stuart,On Saturday, they are creating a Commercial downloadable version of the whole Chat Brain of Open AI. They had to port some stuff to Apache. It is optimized for the M1 chip, but runs on everything. Any kid with a laptop can grab this entire stack. He explains it all and documents it.
In 10 minutes, mega people can download and run this thing. Next is text to video.
I thought this information was absolutely informative.
I think this is a great find, though I don't think many others on here have understand it's significance.
I just got around to watching the video you linked. As you say, they have essentially trained a chatbot similar to ChatGPT which they call GPT4All. They used Meta's Large Language model (similar to GPT3) available for research purposes, then trained it using ChatGPT prompts. End of story, they created a smaller chatbot model that can run on someone's laptop and which randomly enough is a 4-bit model.
Now the limitation with using Meta's model is that it can't legally be used for commercial purposes. So in a few days time they'll be releasing another small model that is using a large Open Source dataset, which can be used for commercial purposes.
Unlike the big companies, they are releasing these models for everyone to use. Not only that, but they're releasing the instructions so that anyone can create their own models that can be used for commercial purposes.
This is just the start of almost anyone being able to make their own Chatbot models for business opportunities. And as the guy was saying on the video, the Chatbot they produce could potentially be placed on something small like a Raspberry Pi. Others in the machine learning community will eventually improve on these models and make them smaller and better. So hobbyists and the like will now have their own way of making their own versions of this.
There's a good chance that big companies are already talking to Brainchip about using Akida for these purposes. But what this open sourcing of chatbots will do is enable more custom and innovative AI applications that will continue to grow. Since it can be done for less than $1000, a lot of people will be able to create their own chatbots, it will no longer be limited to just the companies with millions of dollars.
If you were going to run a small chatbot on something like a Raspberry Pi you'd want it to have an efficient, high performance machine learning chip ideally so it wouldn't run too slow. This is where I think Akida is in the right place at the right time.
Pure speculation, DYOR
Bravo
If ARM was an arm, BRN would be its biceps💪!
Rob “likes” something from someone partnered with Qualcomm... I wonder what that means?
Qualcomm Expands Connected Intelligent Edge Ecosystem Through… | Amit Mate
GMAC Intelligence is excited to partner with Qualcomm as they expand the Connected Intelligent Edge Ecosystem through groundbreaking IoT and Robotics products.…www.linkedin.com
View attachment 33424
I'm bologna way by your by-line.Rob “likes” something from someone partnered with Qualcomm... I wonder what that means?
Similar threads
