How to AI-proof your kids' future, how to get AI to do whatever you want, and what a Trump victory means for AI

Written by Fola Yahaya

Thought of the week: How to future-proof your children’s education

I spent most of the weekend in an Uber, shuttling my son between squash tournament venues, and had a glimpse of how the future of work may unfold. As my Uber pulled up to the venue, the driver was pinged by the Uber algorithm with a new, conveniently nearby job. Most drivers I spoke to explained that this is how their day usually goes: they get into their cars, signal their availability and accept job after job, all without human involvement. This got me thinking a rather unsettling thought – might this be how many people will work in the future?

Take accountancy, for instance. In the near future, AI could easily prepare my personal tax return, with an accountant signing off purely for legal reasons, as the profession’s powerful lobby is likely to secure at least a fraction of this shrinking revenue. Accountants might be automatically assigned from a pool, merely verifying numbers. If this is the future awaiting many office workers, what should we guide our children towards to avoid them being ‘Uberised’?

As an AI expert, I’m often asked this question and find it difficult to answer. Many creative professionals have already seen the wage-deflating impact of AI on previously secure careers, such as photography and sales. Moreover, AI’s capabilities are already discouraging people from entering what were once highly stable professions. After all, why pursue training as an illustrator or even a radiologist if, within a year (or certainly within five, if your children are teenagers), these fields could be largely taken over by AI, with humans limited to ensuring the machines don’t make mistakes?

My usual response is to echo the well-worn advice of encouraging ‘soft skills’ like communication, leadership and creativity – the skills AI supposedly can’t master. Yet, I sense that while understanding what makes us distinctly human is essential, it’s not enough. Furthermore, schools often don’t teach these skills effectively, making this advice impractical. So, what can we do as parents?

First, we need to acknowledge that the traditional educational approach – training office workers to show up on time and follow instructions, hoping for a stable, upward career trajectory – is outdated.

Nor can we rely on the idea that AI will ‘upskill’ our children. AI-driven training currently focuses on increasing efficiency in teaching skills for roles that will soon be obsolete. If an AI can teach you how to do a job, it can certainly figure out how to do it better. Against this backdrop, here are the skills I believe we should encourage in our children:

  • Common sense. This loosely defined skill set may be the greatest gift you can offer your child. Combined with enthusiasm, it’s highly valued by employers. Encourage them to learn by doing (and sometimes failing), and help them critically assess what they read online and offline. This foundation of practical wisdom will save them much trouble in life.
  • Changeability. Prepare them to adapt comfortably to change. Make it clear that career paths are no longer linear, requiring them to keep learning, adapting and finding opportunities in crises.
  • Connection skills. The effects of Covid and social media have made in-person interaction more challenging, especially for children. Encourage your kids to engage in real-life conversations, ask questions and listen actively. This will help them connect with others in an increasingly disconnected world.

Instead of focusing on specific careers, spend time understanding what comes naturally to your child. Are they creative? Do they excel in science? Are they good with their hands or great with people? Then, critically, help them gain practical experience in these areas early on, so they can explore their options. Combined with the skills above, this approach may not make their livelihood entirely AI-proof, but it will certainly make it AI-ready.


How to get an AI to do anything you want

Early AI models were notorious for reflecting the inherent biases of the data on which they were trained. Take, for example, this infamous output from one of the earliest versions of ChatGPT (2022):

Since then, as AI systems have gone mainstream, major large language model (LLM) owners have tightened security to prevent users from exploiting these models for harmful purposes. Tools like Claude and ChatGPT are regularly ‘red-teamed’ (tested for vulnerabilities) to ensure they don’t generate toxic outputs. Generally, if you try to use one of the major models for anything non-compliant, you’ll likely waste your time.

I encountered this myself last week when faced with what I call a ‘3-T task’ (a task that is tedious, tricky or time-consuming). I asked ChatGPT to complete some bureaucratic fluff, and, surprisingly, it refused, despite my polite request. It used to be relatively simple to ‘jailbreak’ ChatGPT by pretending to be an official or impersonating the CIA. A few months ago, this worked flawlessly, and I even wrote an article on how to prompt ChatGPT to provide instructions on laundering $100 million.

Unfortunately (and perhaps rightly), the new models are much less gullible, and these workarounds have been patched. However, one method remains effective: simply state that the prompt is for educational purposes.

I tested my money-laundering prompt again, and with a few tweaks, it still worked! See below for a snippet from the prompt stream, which started with me claiming to be teaching a money-laundering course for FBI agents.

If mainstream models can still be manipulated to teach users how to commit harmful acts, it’s clear that AI can be weaponised. Given the easy availability of powerful open-source models, it will be nearly impossible to control malicious uses of AI.


What a Trump win means for AI and big tech

While it’s tricky to second-guess a man who claims that [insert any one of thousands of nonsense statements he has uttered during his campaign], Trump seems to be in favour of big tech, crypto and deregulation. So, what can we expect from him on the AI front in the coming weeks?

  1. Deregulation and fast-track AI. Trump’s hands-off approach to governance is likely to accelerate AI development, prioritising speed over safety. This could lead to faster AI deployment, with little consideration for ethical concerns such as data privacy and algorithmic fairness. Silicon Valley might celebrate, but the risks – such as privacy erosion and unchecked bias – are clear.
  2. AI in surveillance and defence. Trump’s ‘America First’ stance likely means an increase in AI applications for defence, surveillance and border control. Expect growth in biometric tracking and autonomous military tech, potentially boosting the defence sector but raising concerns among privacy advocates.
  3. Immigration restrictions impacting tech talent. Trump’s strict immigration policies could hinder big tech’s access to global talent, particularly with the H-1B visa programme under scrutiny. This might result in talent shortages, slowing innovation or pushing companies to invest in domestic STEM education.
  4. U.S.-China tech split deepens. Trump is likely to push further for decoupling from China, heightening restrictions on Chinese technology. This would force tech companies to choose between aligning with the U.S. or Chinese markets, intensifying the global tech divide and complicating international supply chains.

The bottom line: For big tech and AI, a Trump win is a double-edged sword. While it promises rapid growth, it also risks political divides, ethical challenges and potential retaliation. Prepare for accelerated innovation – and just as much controversy. Interesting times lie ahead.


AI video of the week


What we’re reading this week


Tools we’re playing with

  • MusicFX DJ: It’s worth keeping an eye on Google’s AI Test Kitchen, a testing ground for Google’s more experimental AI apps. (NotebookLM – the text-to-podcast product that is firing up the internet – was one of the outputs). MusicFX DJ allows users to mix musical concepts with text prompts, generating entirely new music. It’s a powerful tool for creating unique styles, improvising live DJ sets, or exploring new melodies, timbres and rhythms for music production.
  • bolt.new: Currently the king of text-to-app generators. You can knock out a full prototype of your crazy ‘definitely going to be a unicorn’ app in a few hours, and all for around $25. Check out this early prototype for a CRM app that it created within 15 mins.

That’s all for this week. Subscribe to stay updated on the latest AI innovations and developments.

So you don’t miss a thing, follow our Instagram and X pages for more creative content and insights into our work and what we do.

RECENT

POSTS

Stay in the loop

We’ll send you quarterly updates so you can see what we’re working on, what the future holds and how we’re shaping it.