[hamdani yusuf (OP)]

The code for AGI will be simple | John Carmack and Lex Fridman

[hamdani yusuf (OP)]

Another good news in virtualization.

Unreal Engine 5.2: Incredible Simulations!

[hamdani yusuf (OP)]

I just asked Google Bard to translate an e-mail written in German into English. Somehow it translated the e-mail into Thai.
My first prompt contained both the command to translate and the quote of the e-mail.
I prompted it again with only the e-mail quote, and it replied with Thai translation again.
I finally get the correct reply when I prompted again, this time with the command only.

It seems like the AI has gotten more "creative". It shows that it still has room for improvement.

[hamdani yusuf (OP)]

NVIDIA's New AI: Text To Image Supercharged!

The improvement in efficiency here is staggering.

[hamdani yusuf (OP)]

Musk: Tesla FSD Beta Has A MIND!! What Does This Mean?!

[alancalverd]

It means that Elon Musk is liable for any and all  accidents involving the product.

[hamdani yusuf (OP)]

It means that Elon Musk is liable for any and all  accidents involving the product.

It depends on terms and conditions.
Beta version leaves responsibility to the users.

[alancalverd]

You mean like a car that you drive?

The rule in aviation is that the pilot is entirely responsible for what he does (except when under acknowledged direct control by an authorised air traffic control officer) but can accept information and guidance from others - including machines. There is a clear distinction between "be advised that..." and "turn left immediately". You can override a GPS-coupled autopilot and even the automatic trim (thanks to the 737MAX debacle).

My bog-standard car gives me lots of information about speed,  temperature, fuel state, position, speed limits, even traffic priorities, but my mind makes all the decisions - even to ignore warning lights or traffic signals if some greater emergency turns up.

So the question is just how much autonomy  does the Tesla have? Anything that can't be overridden is the liability of the manufacturer - you can't exclude fundamental product safety liability through Ts and Cs.

And there is a very strong legal precedent. Back in the 1960s a small company manufactured an intrauterne contraceptive device that did a lot of harm. By the time a class action was initiated for compensation, the company was bankrupt and dissolved. So the courts agreed that the victims could sue Dupont Chemicals, who supplied the raw plastic material, even though they had no part in the design or clinical trial of the device. Massive payout, and it is now impossible to buy any Dupont product for medical use, which makes thick-film circuits difficult to incorporate in a medical device since Dupont make really good TF inks!

[paul cotter]

Same with any components( principally semiconductors ) I have bought for quite some time now. "Not to be used in any medical device" or "not to be used in any medical device without written permission", company shall not be liable etc,etc.

[hamdani yusuf (OP)]

Anything that can't be overridden is the liability of the manufacturer

I never drive a Tesla, but I've watched YouTube videos by Tesla owners. They mostly give positive reviews.
Afaik, Tesla FSD can be overridden anytime by the driver, as shown in many earlier videos. But lately, there are more videos showing FSD journey fully automatically with no intervention.

[hamdani yusuf (OP)]

How AI Unlocks Hidden Insights in Research Reports

Unlock Hidden Insights in Analyst Research Reports with AI

Analyst reports contain a goldmine of market intelligence, but key insights are often buried across hundreds of pages. Reading these dense reports to find relevant information is incredibly inefficient.

Now, innovative technologies like vector search engines, machine learning algorithms, and natural language processing are transforming how insights can be extracted from research reports.

See how vector similarity models like Pinecone and FAISS convert unstructured text into structured vector data optimized for semantic search. Queries based on contextual meaning are now possible, beyond just keywords.

Large language models like GPT-3 and Claude analyze query context and deliver concise answers drawn from connected insights across sources. Reports become interactive portals instead of isolated documents.

This video explores how AI is revolutionizing business intelligence extraction:

Vector search vs traditional keyword search
Semantic similarity and relationship understanding
Automated synthesis of insights across reports
Conversational interfaces and natural language processing
Increased efficiency and relevance
Discover how technologies like vector databases, machine learning, and chatbots can unlock hidden insights in analyst research reports. The future possibilities for leveraging AI to enhance business intelligence are limitless.

0:00 - Introduction

0:23 - The problem with analyst research reports
1:35 - Vector search engines explained
2:02 - How vector similarity works
2:48 - Converting text to vectors
3:17 - App demo - Future of AI
4:42 - Tailored answers
5:05 - App demo 2 - Macroeconomics
6:45 - The power of semantic search
7:57 - Shameless self-promotion
8:40 - Outro (those bricks again...)

[alancalverd]

Analyst reports contain a goldmine of market intelligence, but key insights are often buried across hundreds of pages. Reading these dense reports to find relevant information is incredibly inefficient.

Which is why reports always have an executive summary or abstract at the top, and conclusions and recommendations at the end. Nobody reads the bulk of the text unless these "insights" are really important.

In the bad old days of print, I worked with a civil service translator on a number of Japanese and Russian papers. He scanned titles for stuff relevant to his clients, and if we wanted more he simply translated the abstract and the axes of the graphs. If that was really interesting, he'd read the whole paper to you but that was very rarely necessary because we had Real Intelligence.

[hamdani yusuf (OP)]

we had Real Intelligence.

What's the main distinction between your real intelligence and AI?
Do you think that future AI can have it as well?

[hamdani yusuf (OP)]

But lately, there are more videos showing FSD journey fully automatically with no intervention.

Musk: GOODBYE to Legacy Tesla FSD CODE!! Plus, Is Compression Intelligence??

In a recent post, Elon Musk says that the Tesla AI team is training a NEW FSD Beta architecture that is replacing about 300,000 lines of legacy code with only 3,000 lines of Neural Network, Software 2.0 code--and that this is the final piece of old software 1.0 to go! Not only is this exciting news but it ties into a theory that says compression is intelligence, which would mean the new code is a whole lot smarter than the old!

[hamdani yusuf (OP)]

There are too many articles about AI we can read with our available time. So, here are some headlines.

11 Major AI Developments: RT-2 to '100X GPT-4'

There were 11 major developments in AI in just this week, from RT-2 ? a significant step on the path to robotic AGI ? to ?100x GPT 4 in 18 months?, and from the uplifting news of TranscribeGlass and AI Barbie-heimer to dramatic revelations about OpenAI in the Atlantic.

We?ll also glimpse Stable Beluga 2 ? the ?first true competitor? to ChatGPT, based on the open source Llama 2, hear about Universal Jailbreaks, learn that OpenAI has surrendered on Text Detection, plus I?ll cover the highlights of the Senate hearing on AI.

Chapters:
0:18 ? RT-2
2:46 ? 100X GPT-4
3:57 ? AI Video
4:29 ? Altman Atlantic
8:41 ? Jan Leike Interview
10:02 ? Speech Transcription + Generation
11:07 ? OpenAI Text Surrender
11:43 ? Stable Beluga 2
12:51 ? Universal Jailbreaks
14:19 ? Senate testimony: Bio
16:59 ? Senate Testimony: Security

[hamdani yusuf (OP)]

We've seen a lot of AI news talking about high level software, which are more focus on the application of AI models to solve users' problems. But those AI models need appropriate hardware and low level software to run the system for training and deployment.

NVIDIA's GPU Boom Explained. Are There Any Worthy Alternatives? Tesla, Cerebras, AMD

This New AI Supercomputer Outperforms NVIDIA! (with CEO Andrew Feldman)

In this video I discuss New Cerebras Supercomputer with Cerebras's CEO Andrew Feldman.
Timestamps:
00:00 - Introduction
02:15 - Why such a HUGE Chip?
02:37 - New AI Supercomputer Explained
04:06 - Main Architectural Advantage
05:47 - Software Stack NVIDIA CUDA vs Cerebras
06:55 - Costs
07:51 - Key Applications & Customers
09:48 - Next Generation - WSE3
10:27 - NVIDIA vs Cerebras Comparison

[alancalverd]

What's the main distinction between your real intelligence and AI?
Do you think that future AI can have it as well?

The ability to surprise me.
No.

[hamdani yusuf (OP)]

What's the main distinction between your real intelligence and AI?
Do you think that future AI can have it as well?

The ability to surprise me.
No.

So, you set yourself as a standard?
How reliable is it?
What's so special about you that future AI can never emulate?

[hamdani yusuf (OP)]

NVIDIA Omniverse: Virtual Worlds Come Alive!

[hamdani yusuf (OP)]

How do you define learning?

Better yet, you can watch Neural Networks Learning in this video.

Timestamps
(0:00) Functions Describe the World
(3:15) Neural Architecture
(5:35) Higher Dimensions
(11:55) Taylor Series
(15:20) Fourier Series
(21:25) The Real World
(24:32) An Open Challenge

It's important to note that human brains are neural networks.  Their sheer size and complexity might have caused some people to invoke supernatural explanation just to explain how they work.