Are Large Language Models Really AI?

Kinda awkward seeing the guy presenting how this development is basically a normal agile routine, but with agents instead of real people.

And it'll have the same issues Agile has, it's a useless model that just seeks to check boxes to justify funding by sending weekly status reports to mid level management.

nobody knows what the *** "intelligence" actually means.

(perhaps reductive)

"Does this system actually have AGI if it can't provide a historical materialist analysis of Zoroastrianism on the world stage?" I dunno, can you?

> I need a script that rename all images with only numbers. There you have.

> I need a script that resize all images to something close to 1920x1080. Ask no more.

> I need a script that collect all images and build a PDF. Done

> I need a script that improve the image quality of those images without any human intevention. Easy peasy.

Definitely reductive, since we nearly fully understand how the human brain works, especially when it comes to creating ideas, understanding mathematical concepts like Gödel’s Theorem, etc.

I don't have to. Only one member of our species has to.

All of this code already exists out there on the internet. "AI" is just giving it to you, acting basically like a very expensive search engine and not much else.

Also because its training is based off Reddit and stackoverflow, it will also be snarky and patronising at you just like a real human nerd.

The company whose ‘AI’ was actually 700 humans in India
Disaster as Microsoft-backed unicorn implodes.

Red faces and recriminations are flying thick and fast after revelations that UK AI company builder.ai has been tricking customers and investors for eight years – selling an advanced code-writing AI that, it turns out, is actually an Indian software farm employing 700 human developers.

Founded in 2016, the company, also known as Engineer.ai, became a darling of the AI space and marketed itself so well that it attracted $700 million ($US444.5 million) investment from the likes of SoftBank, Qatar, and Microsoft – which integrated Builder.ai into its Teams collaboration suite.

Builder.ai CEO Sachin Dev Duggal built the house of cards based on the purported capabilities of the company’s AI assistant ‘Natasha’, which was said to be a no-code tool that could build apps six times faster than traditional development processes and be 70 per cent cheaper.

Duggal became an AI industry stalwart, giving himself the title of ‘chief wizard’ at the company and appearing as an expert on podcasts and news outlets that lauded him as a captain of the emerging AI industry.

Yet it all came undone after revelations the company – which appointed new CEO Manpreet Ratia in February – will declare bankruptcy after major backer Viola Credit demanded immediate repayment of the $77 million ($US50 million) loan it extended to Builder.ai in 2023, when it was valued at $2.3 billion ($US1.5 billion).

“Despite the tireless efforts of our current team and exploring every possible option,” the company said, “the business has been unable to recover from historic challenges and past decisions that placed significant strain on its financial position.”
Pulling back the curtain

The bankruptcy came after a Bloomberg investigation found that the company had been working with Indian social media startup VerSe Innovation to undertake questionable financial practices – including using ‘roundtripping’ to overstate revenues for 2023 and 2024 by a factor of four.

As well as its financial shenanigans, the company’s questionable business has been outed after confirmation that Natasha was not, in fact, an AI at all – but a team of 700 very human Indian developers who were not only writing customers’ software, but tasked with behaving like bots.

Rumours to this effect had circulated since a 2019 Wall Street Journal report questioned Engineer.ai’s bona fides, but the truth was finally exposed this month with confirmation that the entire company was a lie – with former employees describing the company as “all engineer, no AI.”

Although the developers used a range of software tools in their work, coding was performed manually, meaning that while Builder.ai did eventually deliver apps to its customers, it was simply another player in an Indian offshoring industry attracting $27 billion ($US17.7 billion) annually.

That puts the company in a completely different market segment than the one that propelled AI-hungry investors through four funding rounds before and after the debut of OpenAI’s ChatGPT turned the global tech industry on its head.

A reckoning for the AI industry

The combination of financial malfeasance and outright deception has reportedly attracted the attention of US federal prosecutors, with suggestions that a formal criminal investigation has kicked off as Builder.ai careens into the annals of AI’s worst ever disasters.

Reports suggest that around 90 per cent of AI startups fail within a year, whether due to a lack of market demand, financial instability, operational challenges, or AI’s technological complexity – and the demise of the likes of Artifact, Shyp, Tally, Eaze, and Ghost Autonomy are prima facie evidence.

Even major firms have struggled to get AI right, with McDonalds, Air Canada, Sports Illustrated, iTutor Group, and Zillow among those that have backtracked after wrongfooted AI implementations.

Gaps between AI hype and AI reality are regularly proving disastrous – as when the surprise debut of Chinese generative AI venture DeepSeek sent share markets tumbling in January, or when Microsoft and AWS announced in May that they would pause their expansion of AI data centres.

As of press time, Builder.AI’s website still claims that “AI means we can build more cost effectively and at speed”, describing an AI-based workflow in which Natasha is an “AI project manager” that helps customers “order an app, like you would a pizza.”

stack overflow

The "AI" is just sorting through all the human provided examples and weighting them by popularity as part of it's training.

Also because its training is based off Reddit and stackoverflow, it will also be snarky and patronising at you just like a real human nerd.

bad dog, don't personify the LLM

This is patently untrue--- at least in the sense that is relevant to this discussion. There is solid consensus in areas like neurochemistry and neuroanatomy, as well as (broadly) psychiatry and how the peripheral nervous system works. Like, the "medical/clinical" aspects of neuroscience. But for the discussion, the key relevant questions pertain to how the brain and the rest of the nervous system gives rise to emergent qualities like "the mind" and "experience." We don't even have consensus on neural coding. There is no existing coherent model for how the human mind works, much less how the brain helps manifest it. Questions very much related to "intelligence."

A good example of this is the NHS in the UK, they import cheap and under-educated doctors, nurses and carers from the 3rd world (often with fake or inferior degrees) and since you have no option, you just have to put up with these people treating you. They don't even want to take on young British doctors and nurses anymore who they trained, cause are more likely to unionize and demand more pay rises from the government. So these people are often forced to goto the US to find work, due to demanded DEI hire quotas at home that allow them to hire these cheap immigrants over them. There are no jobs for the home grown doctors and nurses they trained, even though they say there is a shortage and that is why they have to import.

There are propaganda posters in any health location, telling the UK people that diversity doctors and nurses from the 3rd world are saving the NHS (even though they can barely speak english and are trained to a laughably low level). What they mean is, saving the government who pay them less and don't have to put up with them joining a union.

Big employers (also the government) have never liked employing entitled first world workers, even if they are more skilled they don't like them cause they are entitled (take time off to have a holiday, want a social life, want more money inline with inflation, want to join a union, want to go home and sleep in their bed instead of sleeping on the floor in the office).

Never before in our history has it been so easy to import millions of people who will not (and cannot) stand up for themselves to replace you, and give them a laptop with chatgpt that is trained on your work to fill in the gaps.

Reasonably important stuff

Which is totally illegal but never, to the best of my knowledge, enforced.

Gee golly wiz, just so amazing how that coincidental that is. Doesn't help that these companies donate huge sums of money into the election coffers and book deals of various politicians and government executives.