Team-BHP (https://www.team-bhp.com/forum/)

-   Shifting gears (https://www.team-bhp.com/forum/shifting-gears/)

-   -   The Generative AI Thread (chatbot, text to image, text to video etc) (https://www.team-bhp.com/forum/shifting-gears/259867-generative-ai-thread-chatbot-text-image-text-video-etc-12.html)

I recently had an interesting encounter with Google's Bard AI that revealed an intriguing quirk worth discussing. While Bard AI is widely acclaimed for its poetic prowess and storytelling abilities, it seems to stumble on a simple arithmetic calculation.

During a conversation, I asked Bard AI to compute a life path number for a specific date of birth. Surprisingly, Bard provided a numerical calculation that resulted in an incorrect answer, showcasing a simple arithmetic mistake.

Have any of you experienced similar quirks or surprises with Bard AI or other AI systems? Let's delve into these curious AI traits and share our thoughts!

And it has started happening.

TCS bets big on Azure Open AI, plans to train 25,000 engineers

According to Stability AI CEO Emad Mostaque, most outsourced coders in India will lose their jobs in two years

https://cio.economictimes.indiatimes...ders/101915631

Well, as I have said before, it is not going to happen in near future. These experts who are making these predictions, have no idea about software services industry. If you try to provide software services using AI, there will a whole bunch of very angry customers.

This is turning out to be like Greta Thunberg's predictions on climate change:

"Most outsourced programmers in India will see their jobs wiped out in the next year or two, Stability AI CEO Emad Mostaque said."

"Mostaque, on a call with UBS analysts, said that most of the country’s outsourced coders will lose their jobs as the effects of AI mean that it is now possible for software to be developed with far fewer people."

Link

Has any assessment been done in terms of carbon emission comparison of the same job performed by AI vs human being, on standalone basis (cost of energy and resources for specific task) as well as overall infrastructure (cost of "living" also included)?

I don't see that happening anytime soon. Some low end tasks yes, in bits and pieces yes but not bespoke jobs. I can't say where things will be in 20 years but not anytime soon. As an AI CEO this is the sort of eye catching statement he needs to make! Also demand for software continues to grow at a very rapid pace as digitization enters all aspects of our lives. Need for software over next 25 years I suspect is a multiple of the software capital built over the last 25 years.

As part of a bigger task, the team I work with was supposed to provide some work last week. No one got around to doing it since all of us were busy with our segments of the bigger task. So the manager that was waiting for the overall output used ChatGPT to generate that tiny bit.

The generated output was not as elegant as one of us would do it, in fact, it was a bit hollow- but it did the job. And since the manager had 'generated' it, it went out with the whole bit of work (that team has a HiPPO problem). Now no alarm bells yet- what was handled was a tiny bit of the overall work and the rest of the work can't be handled (yet!) by LLMs. But the writing is on the wall. Better we start using these tools for those bits of work, and gain efficiencies, before we have no say in the matter at all.

Marketing doc.

What kind of work was it? Documentation, code, test plan, design, ...

Of course CEO is playing to the gallery for news, however :
AI will increase productivity of software development for sure. As a practical example :

1. Take a DB Dump of TeamBhp-like Forum
2. Configure a pipeline with OpenAI / Palm / LLaMA to text to SQL
3. Run queries like : "Who is biggest evangelist for Maruti" / "Who is most popular Author" / Which hour of day is busiest for XYZ Topic

I think most developers will be surprised at the quality of Output and ability to make sense of a DB dump and figure out context.

Ah, very easy to do with that since probabilistic output is acceptable for marketing materials.

If the output is supposed to be deterministic, then it wouldn't work.

These GenAI models are getting pretty darn good. Very soon these capabilities will be embedded within Microsoft Office as Copilot. I am waiting for the day when those powerpoint slides will get automatically created just at the click of a few buttons. These tools are going to be the biggest productivity hacks of all time and people who are familiar with it will definitely have an advantage.

Just look at this 2 minute video as an example:
https://www.youtube.com/watch?v=S7xTBa93TX8

Oh already there to some extent- what I remember the CEO at my first job in the early 2000s asking a graphic designer to do to make his important slides more visually appealing, is already being done within PPT by the 'Designer' feature. (And a lot of vacuous, higher-management presentations are basically 90% fluff anyway- so those are easily LLM'd :))

Yep, that's what I was getting at, better to adopt and use as productivity hacks instead of worrying if they will replace humans, etc.

Same thing was said about DIY software such as Foxpro / Oracle Forms in 90s.. I.e. developers will not be needed , just that developers got way more productive once routine tasks were automated away.

As you said, it is crucial for developers to learn these tools to remain relevant few years down the line.

We are using AI tools for proposal making at a client's space. It is not client ready, but it provides a very good first draft to then work on, referencing past tenders for components used, costs etc.

In other words, a more sophisticated "mother" proposal -- if you understand what I mean...