AI is improving rapidly (CHATGPT)

[SOL]

Here's an excerpt from a newsletter I received today (The Tokenist 5 Minute Finance, free):

Over the past couple months, ChatGPT users might've observed an uptick in the model's hallucinations and a drop in reliability. This is what typically happens when large language models (LLMs) can’t tell the difference between fiction and reality.

Further, developers could aggravate the problem by making the AI too accommodating to user’s input expectations. The more the public interacts, the more chances for errors - earning it unflattering labels.


Across the board, GPT-4 took a drastic performance blow. Image courtesy of Stanford University and UC Berkeley.

The AI, once renowned for pioneering LLMs and experiencing record user growth, now fumbles with logic.

For example, even if it correctly deconstructions a problem, it might conclude inaccurately like asserting “17077 is a prime number”.

Two implications emerge:

• Warning Enchantment with AI: The initial awe over AI's capabilities is now giving way to frustration. This raises concerns over its potential integration into business operations.

• AI's Nascent Stage: Despite multi-billion dollar AI investments, we are just entering the AI race.

OpenAI may have sparked the pioneering push, but if not careful, it may end up like Napster, forgotten in the bin of internet history.

To that effect, it is no wonder why Meta and Microsoft, both traditional monopolies, are willing to explore an open-sourced approach to LLMs. This is the Llama 2 family of LLMs, now powered by Microsoft’s Azure computing cloud.

Their objective is refining AI while managing the infrastructure efficiently.

Meanwhile, Apple is experimenting with 'Apple GPT' in its Ajax framework, inspired by Google's Jax. Apple GPT launch is eagerly anticipated next year.

On Google's end, the rumor mill turned into high gear as well.

According to NYT’s anonymous sources, the data monopolist has approached Times, Washington Post, and Wall Street Journal to test a news-generating AI suite called Genesis, which aims to revolutionize content creation.

Ultimately, the AI that hallucinates the least will be the victor.

Full study referenced here: https://arxiv.org/pdf/2307.09009.pdf

The Napster analogy is perfect. The article hints that companies embraced the AI revolution too hastily. Stage 1 is not the phase where you pour billions; it's the stage where observation is key, as approximately 90% of these players tend to vanish. Stage two presents better odds: around 50% will survive, but perhaps only 10% will truly thrive

[AstuteShift]

Out of all the companies, Meta actually has a good position in terms of its AI, I can definitely see them as a potential winner in the long run

Zuckerberg, I despise this guy fro destroying society or helping it to its destruction but that’s what makes it an even better investment

They are still doubling down on the meta verse. Zuck is ruthless

[SOL]

Shifting gears, a fleet of attractive robots is en route, poised to assist you in acquiring even more items that might not be essential.

https://youtu.be/53Pr1FTxjQk

Here's a glimpse of AI's impending dominance:

Artificial Intelligence stands on the cusp of assuming control over diverse facets of our lives. Given the swift technological strides, AI is poised to transform healthcare, finance, transportation, and beyond sectors. It will elevate decision-making, streamline repetitive tasks, and provide unprecedented personalized encounters. As AI advances, its capacity to analyze extensive datasets and derive insights will yield enhanced predictions and solutions. However, for the moment, it's kicking the humans but in in ping pong.

https://www.youtube.com/shorts/nmHzvQr3 ... ture=share

[harryg]

That is a fake video, this is the real one... or is it? Who can tell nowadays?

https://www.youtube.com/watch?v=yn155lSJ5qY

[nicolas]

Groq (https://groq.com/) now gives us access to Llama3-70b, the open-source model from META that is said to rival the most advanced commercial models like GPT-4.

[deepthinker]

And now we have GPT 4o. I was just messing around with it. Its a vast improvement of GPT4. I watched a video of two AI's chatting, they actually sounded more coherent/smarter than many individuals I know

[Budge]

And now we have GPT 4o. I was just messing around with it. Its a vast improvement of GPT4. I watched a video of two AI's chatting, they actually sounded more coherent/smarter than many individuals I know

If it’s true that our species is alone in the universe, then I’d have to say that the universe aimed rather low and settled for very little. –George Carlin

Appears GPT will help us out.