DeepSeek, Cheat Sheets & Use Cases
5 in 5 - Brave & Heart HeartBeat #226 ❤️
This week we’re doing a deep dive on the latest in AI, from DeepSeek’s appearance on the market, how people who aren’t into tech use AI and a very useful AI cheat sheet – you’re welcome.
Plus, what are the actually useful ways to use AI and why are DEI measures suddenly the enemy?
Let’s get into it.
Were you forwarded this? Not a subscriber? 👉 Sign up here
#1 - DeepSeek – The Empire Strikes Back
So, the name is terrible, but we would hate to be on either the trading deck or OpenAI’s offices as the newest AI venture from China has exploded the stock market and stuck a rocket up the behind of American AI ventures.
DeepSeek claims to have created an AI to (almost) rival OpenAI’s ChatGPT using fewer chips and cheaper less powerful chips.
This raises two questions.
One, has the US multibillion dollar AI spending spree really been worth it, if it can be done cheaper? And two, have the US done enough to stop China getting ahead in AI.
Because although US sanctions have banned Nvidia’s most advanced chips, H100s, they’ve managed to get it done anyway using the less advanced technology of the H800 chips. They’ve also been banned, by the way, since October, but too little too late, it seems.
Described as a “Sputnik moment” by experts, it’s also come at an impressively ironic time as the US try and ensure TikTok is free from Chinese influence.
It also seems to demonstrate that you can do innovative things with AI without having the massive budget of Sam Altman.
#2 - The Less We Know The Better
Who do you think is more likely to want to use AI technology, the tech savvy or the slightly less so?
Well, the answers may surprise you, as it turns out that recent research has found that people with less knowledge about AI are actually more open to using it in their daily lives.
People in countries with lower average AI literacy are more receptive to AI adoption, and a survey of US undergrad students shows those with less understanding around AI are more likely to use it for their assignments. This is referred to as the “lower literacy-higher receptivity” link
Apparently the reason behind this is linked to the “human” aspect of certain AI tasks.
People who understand less about how AI works are more impressed by the “magical” nature of AI creating traditionally human outputs like music, art or writing.
Those of us with more knowledge about how AI works see the technology as less mysterious, and in contrast are more focused on AI’s efficiency and prefer tasks that don’t evoke human qualities, for example analyzing test results.
AI is everywhere, do you have the vocabulary to talk to people about it?
Well, you do now.
Let’s start with the basics.
Generative AI: AI technology which can generate text, images, code and more after being trained on a LOT of data. Think ChatGPT or Gemini.
AGI (Artificial General Intelligence): AI that’s as smart, or smarter than a human. Bascially mega AI – the stuff you see in the movies. See also, “superintelligence”, AI that is MUCH smarter than a human.
Hallucinations: This is when generative AI tools make up, with confidence, absolute nonsense that they think is the right answer to your question. Think the disastrous Google AI which told us all to eat rocks.
Bias: A huge problem in AI – tools can demonstrate bias depending on what data they’ve been trained on. They basically “think” whatever the data they’ve learned from tells them to. Think sexism, racism, the lot.
Now a bit more detail. What are the different models?
LLM (Large Language Model): A kind of AI that can process and generate natural language text. That’s what you’re interacting with when you talk to ChatGPT.
Diffusion models: AI models that generate things from text, like images, audio and video.
Foundation models: Generative AI models are trained on a huge amount of data, so they can be used as the foundation for many different applications. For example, GPT, the foundation model behind ChatGPT.
Frontier Models: Basically, stuff that doesn’t exist yet, but with a sexier title. Frontier models are the models that AI companies are working on, but not finished yet.
And some bonus words?
Inference: The term for when a gen AI application actually generates something, i.e. that recipe you asked it to come up with.
Neural network: The architecture that computers use to process data. It can be compared to a human’s brain neurons, important for AI because it means systems can learn to understand complex patterns without programming.
Ready to shine in tech society?
#4 - What’s The Point Of All This Anyway?
We’re all about AI right now, but what can it actually do, apart from writing your homework in your place or generating offensive images?
You may or may not be surprised to know that those are the least interesting and useful things that AI can do. Here are some of the ways AI is revolutionising different sectors:
Aviation: Ryanair use AI to optimise revenues, schedules and pick the best aircraft for each flight. BA use it at to help pick gates at Heathrow according to how many passengers are transferring, and Easyjet use predictive AI technology to redesign maintenance for stock to make sure it’s always up to date.
Energy: Google have used AI since 2019 to increase the accuracy of power output forecasrs for their renewable fleet, which has meant they’ve been able to balance consumption and even sell electricity back to the grid, increasing the value of their wind power by 20%.
Pharmaceuticals: A treatment for a deadly lung disease is the first of it’s kind to be developed using AI to generate 30,000 novel small molecules and then whittle them down to the six most promising drugs, and AstraZeneca say more than 85% of their small molecule drug pipeline is AI-assisted.
Still impressed by ChatGPT?
#5 - What’s The Problem With DEI Anyway?
Basically, it’s woke.
Since the 2023 ruling against race-conscious admission programmes at US universities DEI has been under fire, and since Trump’s second rise to power, companies have really let loose when it comes to axing DEI initiatives.
The battle against DEI is another facet of the conservative battle against wokeness – the epitome of which was Florida trying to pass a “Stop Woke” law that banned companies from requiring diversity training.
In the US people are arguing that Diversity, Equity and Inclusion policies are actually discriminatory against non-minority employees.
There’s a great quote summing up DEI from Ken Frazier, the first black CEO of a large pharmaceuticals company:
“At its best, DEI is about developing talent, measuring it in a fair way and finding hidden talent and disadvantaged talent in a world where not everybody has an equal chance to exhibit their abilities.”
And although DEI measures were originally introduced in order to counteract the negative bias faced by minorities, they’ve also largely been proven to be beneficial for businesses.
One study in 2020 showed that out of 1,000 US companies, those with greater ethnic and gender diverse had a 25% probability of higher profits, and McKinsey found in 2022 that diverse, equitable and inclusive companies were better able to respond to challenges, win and keep the best talent and meet the needs of customers.
Sounds like something we should definitely get rid of then… right?
Brave & Heart over and out.
Bonus
Happy Wife…
Happy life, in prison.
A drug lord was arrested in London after years of evading capture in Costa Rica after his wife detailed their entire European trip on Instagram.
As he was arrested at London Bridge, we’re assuming she was posting touristy snaps on her stories seconds before.
To find out more on how you can retain your top talent, or how we can help you with digital solutions to your business and marketing challenges, check out our case studies.
Prefer to listen next time? Check out our podcast!