People always hear about interesting forms of AI like LLMs but maybe less interesting to an individual and more crucial to a large business is constraint optimisation.

Which is actually the field I focused my PhD on.

A big part of constraint optimisation is the various applications #AI can have in the real world, and so I am delighted to see this example where it has been used to simplify order picking 📦
(And it’s more exciting than it sounds, honestly!)

Order picking is one of the most time-consuming parts of any warehouse worker’s job. But Mercadona managed to streamline the process and increase profitability by using AI to achieve the following:

⭐️ Speed up the order-picking process - they optimised the process to achieve maximum efficiency and productivity.
⭐️ Grouping orders into batches - they discovered new AI-powered strategies to efficiently combine orders into batches, leading to faster fulfilment.
⭐️ Overcoming optimisation hurdles - AI was used to help overcome obstacles and find the best configurations through the powerful Variable Neighbourhood Search (VNS) approach.

AI was so effective here because it explored thousands, if not millions, of configurations to try and find the most optimal solutions iteratively. This can lead to huge profitability gains when done right.

Constraint optimisation can be applied to virtually any area, from warehouse picking here or shift scheduling like we do at Fielder
In fact, one of Bill Gates’ earliest programs was a timetabling software too.

If you’re interested in leveraging AI to provide a fast and easy solution to your business, please reach out to me and I’ll see how I can help you.

(And if you want to read more about AI and order picking, check out the full article here: https://lnkd.in/ezjrbxDK)

Something that I think is super underrated when starting a business is the strength of relying on your own skills.

It’s the one thing that allowed us to launch SmartRev Limited so efficiently.

Both my co-founder Alex Stewart and I knew what each other’s strengths were, and how we could complement our skillsets with one another.

Alex is far too good at marketing but has a developer background. I like to think I’m a good developer and have experience in marketing from running startups.

This means we can trust each other to run our part of the show but can still communicate and understand what’s going on overall.

I’ve built the entire codebase for SmartRev Limited myself and thanks to Alex, people love to use it.

Thankfully, because of this, we’re already profitable.

So if you’re on the edge of starting a business but wondering how you can afford to take the jump, consider playing to your own strengths and finding a co-founder whose skills make up for whatever you lack.

My business partner Alex Stewart has managed over $10 million in ad spend for clients…

…Only for a large percentage of the generated leads to quickly end up dropped.

It wasn’t the only problem his clients were facing:

→ One had a sales rep who closed more than two times the number of leads than another
→ Others weren’t able to sustain conversation to generate leads into clients
→ And some just outright didn’t have time to respond to all queries

And even those clients who DID manage to close all of their leads found themselves confronted with a new problem: booking someone on a call is an expensive thing. You need to be available or train one of your sales reps to handle new potential clients.

All this adds up time and money.

The solution?

AI.

Together, we realised AI can bring a lot of consistency and immediacy to the appointment setting process.

From there, AppSetter at SmartRev Limited was born…

It’s similar to how I had the idea of using AI to improve the job search when I launched Fielder 3 years ago.

Business is all about solving problems - and AI is a pioneering way to do that.

Send me a message today if you’re interested in finding out more about AppSetter!

I wouldn’t be surprised if we saw AI become a school subject in 10 years’ time.

With how quickly it’s advancing now, AI is likely to automate many aspects of the education process in the future, but I think it will also make its way onto the curriculum to help today’s children become well versed in the technology of the future.

An AI school subject could cover things like:

✏️ Using AI for innovative problem solving
✏️ Programming and writing code for AI
✏️ Building understanding processes
✏️ Critical media literacy
✏️ Prompt generation
✏️ Machine learning
✏️ Controlling robots

Making technology a priority just as high as reading and mathematics is important to understand how computers work, how they’re designed and how they influence thinking and learning.

With the world becoming increasingly automated, the younger generation need the technical know-how about how to automate processes and plug systems into one another.

I hope we see AI make its way to classrooms soon… what do you think?

Formula 1 is testing out a new AI system.

It aims to help officials check whether a car's wheels entirely cross the white boundary line, with the margin of success in Formula 1 often coming down to tiny measurements of time and distance.

They used Computer Vision technology at the season-closing Abu Dhabi Grand Prix last month, with shape analysis to determine the number of pixels that cross the line at the edge of the track.

They don’t plan to fully automate reviews of track limits breaches but rather significantly reduce the number of potential rule violations that are sent to officials for manual review. It’s a great example of AI being used not as a replacement but a tool to facilitate a human’s job.

And it’s much needed - July's Austrian Grand Prix saw four people having to review around 1,200 potential violations. After some track limit violations went unpunished in October's US Grand Prix, officials knew they had to do something different, and Computer Vision was the answer.

The technology, which has also been used in medicine to help review data from cancer screenings, should reduce the number of possible infringements that officials manually review to around 50 per race.

In my opinion, it’s a great application of AI and shows the potential this technology has in saving time, reducing monotonous tasks and improving accuracy.

I promised full transparency as I built my new start-up SmartRev Limited and our first product, AppSetter.

So here are some recent updates we’ve made to AppSetter:

💡 Knowledge Banks - AppSetter works wonders with just a brief product description, but now you can also upload longer files including product information, frameworks, offers and more. Our agent will sift through them for those unexpected questions or as part of the sales process. Smart, right?

💬 Channel-Specific Responses - We’ve taken AI-generated messaging a step further and can now tailor responses for emails, SMS or Messenger, meaning the proper etiquette can be used for the right channels rather than a template standard across all.

📱Better SMS Handling - Some users gave us great feedback that the AI was writing super long SMS messages, causing them to fail. So we reworked this and have found a way to break down those lengthy messages to make sure they end up delivered!

🛠 Enhanced Request Tracking - Our dashboard now shows incoming workflow requests so you can keep an eye on everything in real-time.

⚡️ Increased Resiliency and Scalability - We’ve boosted our system’s resiience and scalability so you can now send thousands of messages. The only limit you’ll hit is your OpenAI rate limit.

Great things can happen when you listen to user feedback and are not afraid of reworking and experimenting.

I know this isn’t the final fork in the road for some of these aspects though, and further revision will undoubtedly occur!

The AI tools that have changed our startup ⬇️

(Coming from a second-time founder with a PhD in Computer Science!)

1️⃣ Copilot

Copilot has allowed me to be the sole developer for the entire SmartRev code base and shorten proof of concept time from weeks to days. I honestly don’t know where I’d be without it!

2️⃣ Midjourney

This allows us to generate high quality images and logos without having to pay for expensive designers - a lifesaver in those early days of a startup when cash flow is tight.

3️⃣ Perplexity

I tell anyone who’ll listen that Perplexity is my favourite AI tool of the year. It can save you a lot of time from trawling through the Internet to research something or troubleshoot code. Think of it like a better Google. For SmartRev specifically, it helps us get detailed breakdowns of our competitors in seconds.

If you’re a startup founder, this is a list you’ll want to save.

Your life will become a lot easier once you start using AI in the right way.

(Thank me later)

Any other AI tools great for business owners?

A Microsoft app, that I’ve been using for years as a blind user, has just launched on Android.

'Seeing AI,' which first launched on iOS and last week came to Android, uses smartphone cameras and AI-powered narration to help blind and partially sighted users understand more of the world around them.

The latest GPT version, which in some respects is very advanced, detailed and capable, has the unusual difference and advantage that it uses older tech which may otherwise seem obsolete but it comes alive with GPT vision - you simply have to take a picture and in seconds it can be processed.

I use this app for the following:

🔹 An instant text reader for describing what a photo is showing
🔹 Answering questions about a document, which is useful for things like finding the price of an item on a bill or summarising written text
🔹 Scanning barcodes
🔹 Providing product information like the name and details from packaging, which can be useful for things like bleach where you don’t exactly want to figure out which is the right one by opening and smelling them all!
🔹 Identifying currency, colours and brightness
🔹 Interpreting handwritten text in various languages

I’ve posted before about the frustrations of finding accessible technology for blind users like myself, and it’s encouraging to see that #AI is slowly paving a way forwards.

The main reason why I like Seeing AI is because it’s live.

It’s a great app and I would highly recommend it for other blind or partially sighted users!

Last week I attended the AI Advancement Forum: Revolutionising Education at the House of Lords.

This summit brought together leading universities, academics and industry professionals from across the field of AI.

It’s been exciting to see how quickly the AI world is advancing, particularly in the education sector. We discussed how to build artificial intelligence products that can assist teachers without taking over their roles.

It’s exciting to see the government is truly taking AI seriously and placing it at the heart of the education of our next generation. I hope we can get there with Fielder one day and pitch to the House of Lords about using AI for job centres!

Although integrating AI solutions into classrooms is a cool next step, there are also some controversial issues that bring up a moral dilemma in this field - we criticise students for using ChatGPT in their schoolwork yet are building solutions for teachers to grade their work using GPT.

So, should we really be critiquing these students or instead training them to use AI in a future where they’ll need it to survive?

Overall, this was a great event that I attended with my good friend Dr. Salman Valibeik, and I enjoyed taking in the history and heritage of the House of Lords. It’s a place that’s steeped in tradition - we had to leave the room by 5 PM sharp!

Many thanks to all those who hosted the event.

How I built my (second) startup -

A full breakdown and tips I learned from the first!

I recently shared a post here on LinkedIn about the mistakes founders typically make when building a startup.

Now I don’t claim to have learned them all, but it was important to me to stick to my word and be transparent as I aim to build SmartRev Limited in the leanest way possible.

So here’s how we started this time around…

1️⃣ Once we were convinced that it was technically feasible, we shared the idea before committing to any code. Only once we knew there was a demand did we start putting some real time into properly building it.

2️⃣ Instead of putting in our own cash, we found ways to generate cash exclusively for funding the startup. We used our skills and connections to broadcast a two-hour workshop on leveraging AI within your agency, and sold 40 seats at $100 each.

3️⃣ When we had an idea for a feature, we posted about it (on where??), and those that got reactions, we built. Those that didn’t, we reexamined.

4️⃣ We didn’t wait for it to be perfect. This was a big thing I stressed in my last post. And our first version of SmartRev was built together with string and tape. But it proved feasibility and we got a lot of feedback. Version 2 is still a work in progress, but let’s just say the foundations are there now.

5️⃣ One thing I’ve recently been thinking about is something Sam Altman says - he stresses that the best startups build products that people love so much they tell their friends about it, and that’s what we’ve been seeing. Businesses bring other businesses on board. So don’t be afraid to share your startup with the world, even in its early stages.

I still have a long way to go with SmartRev Limited, but at least this time I’ve done it once before so the startup journey feels a little smoother!

Here’s what our roadmap looks like:

It’s never been easier to build your own low-code or no-code tools, thanks to AI.

Alex and I recently built a system that allows you to drop a file into a Google Drive folder and then automatically transcribe, analyse and summarise it - all with zero code.

The applications are endless:
🔹 Copy agents that follow your tone of voice
🔹 Content agents that repurpose your videos into short and long form posts/blogs/emails
🔹 YouTube script analyzer agents that can break down videos into repeatable frameworks
🔹 Accountability agents that can check-in on your clients/customers/team members

We’re looking to help out more communities and get them up to speed on how to leverage AI for themselves with low-code tools.

If your community would benefit from this, leave a comment below with the phrase ‘low code’ or send me a message and I’ll get back to you soon!

An observation - is it just me or are people not really using Microsoft?

When I heard the news about Sam Altman’s dramatic departure from OpenAI, I noticed that Greg Brockman mentioned several times that he and Altman were asked to join a Google Meet to discuss their future at Open AI.

I was wondering, why are they meeting on Google Meets when Microsoft invested $10 billion in Open AI?!

But then I thought about it a little more and realised something:

Almost all enterprise companies build their business stack around Microsoft and they have dominated enterprise, from government to large businesses to charities. And for good reason - they products are “good enough” and are built to be scalable and resilient.

But the startup world seems different.

💡 Microsoft Word? No, we’re all about collaboration on GoogleDocs.
💡 Outlook? No, we use Gmail.
💡 Teams? No, it’s Zoom or Google Meet all the way.

I’ve always wondered this - why is it that startups, my own included, are happy to pay for a plethora of platforms from Slack for chatting, Linear or Clickup for projects, Notion for notes and so on, when theoretically you can get this all under one price from Microsoft 365?

It used to be that these platforms were genuinely better. I used to find Slack much faster than Teams, and the Zoom background noise filtering so much better, too. But that’s not really the case anymore.

And yet, whilst Microsoft will always remain king because they hold all the large enterprise companies, I wonder what the new generation of enterprise companies, born out of today’s startups, will use.

Maybe Microsoft products just seem outdated because they remind us all far too much of school…

Do you still use any Microsoft products?

Elon Musk is set to release his answer to ChatGPT (called 'Grok') sometime this week.

Will it take over?

My answer is no ❌

Grok allows you to choose from two modes, regular or fun, and Musk says it won’t reject controversial questions or prompts that are often dismissed by other generative AI models.

Grok will use the information from X for real-time world news and describes itself as “a state-of-the-art language model that is significantly more powerful, achieving 63.2% on the HumanEval coding task and 73% on MMLU.” According to the website, Grok has also surpassed OpenAI’s ChatGPT-3.5 and the newer personal AI platform Inflection-1.

It’s not Musk’s first foray into AI, though - he was among the investing founders of OpenAI but left back in 2018 over a potential future conflict of interest as Tesla threw more efforts into AI technologies.

Musk has recently criticised OpenAI for bias and complained about how commercialised generative AI is becoming. For him, OpenAI was "a counterweight to Google.” And now he has rolled out his own platform as a counterweight to all other models.

But here’s why I don’t think it will be:

Putting all of the AI models against each other, GPT-4 still leads by quite some margin. GPT 3.5 is a comparatively ‘easier’ benchmark to get to, hence why it has a few rivals. But GPT-4 is still some way ahead.

Many people don’t realise how many manual hours were put in GPT in terms of real human reinforcement feedback, and it’s this process that OpenAI claim is the key thing that makes it superior to any other AI.

Of course I could be wrong, but I have my doubts that Elon Musk and Grok will match the standard set by GPT-4…

What do you think? Could Grok really become the next GPT?

I am delighted to announce the launch of a new venture with Alex Stewart, a friend of mine for 20 years. We are partnering to launch our company SmartRev Limited, and our first product, AppSetter.

It is a huge achievement for me to be starting my second business after so long in academia and among the other hurdles I have faced along the way.

After a few months of building this ‘in secret’ I am so excited to finally be able to share more details about my new venture.

AppSetter uses AI to automate and take over your appointment setting.

❎ No more money wasted on overpriced setters
❎ No more time wasted on micromanaging VAs

Instead, we combine GPT 4 with GoHighLevel integration and our proven sales processes to effortlessly build rapport, qualify, and book appointments with your leads 24/7 - without you having to reply to a single conversation.

On a personal note, I have been focused on real-world applications of AI ever since embarking on my PhD in Constraint Optimisation at University of St Andrews. I first applied this technology to my hiring and staffing platform Fielder and am now excited to revolutionise the world of appointment setting with AppSetter.

Many thanks to my co founder Alex Stewart for being the marketing brains behind AppSetter and trusting in my AI expertise.

But this is just the beginning - over the next few months we’re adding in even more to AppSetter, from prompt generators to sales call analysers.

If you’re interested in automating your lead conversion and appointment setting process visit our landing page (the link is in the comments!) or send me a message to find out more!

The Power of Diffusion Models:
What are they and what do they do?

💡 Diffusion Models are a class of generative AI models that leverage the power of non-equilibrium thermodynamics. In simpler terms, they essentially generate data similar to the data on which they are trained.

⚡️ How do they work?

Here's a quick breakdown:

🔹 Diffusion models introduce Gaussian noise (a type of random noise with a bell-shaped distribution) to the original data during the forward diffusion process.
🔹 Through training, these models learn to recover the original data by reversing the noise removal process in the reverse diffusion phase.
🔹 By effectively manipulating the noise levels in both directions, diffusion models help create striking visual outputs, often exceeding the limitations of previous generative AI models.

Embracing the potential of diffusion models is crucial in AI.

Pioneers like OpenAI (with models such as GLIDE, DALLE-3) and Google (with Imagen) continue to push the boundaries, so we can expect game-changing advances in image generation and other domains soon.

If you’re interested in learning more about diffusion models, check out the blog post in the comments or send me a message!

#AI #GenerativeAI #DiffusionModels #MachineLearning

For an idea to be successful, it has to be original
Especially in technology…

Wrong ❌

This might be a controversial opinion, but I don’t think you need to reinvent the wheel to succeed. There’s a lot of value in seeing something that already works and applying that in a new way.

For example, the idea for Fielder originated from a project called that used AI to optimise the whole stack - everything from supply chain to drop routing and delivery. It was efficient, innovative and most importantly, worked.

So I knew that if I extended that process to the industry of working people, it would, theoretically, also work.

Creating something from scratch is incredible.

But so is applying an existing technology, process or experience to your industry or audience.

The focus of my PhD, constraint optimisation, was actually about applying different technology to new industry. So I thought that if you can do it for scheduling machines, why not people too?

We only have to look at all the different innovations people are using GPT for to see evidence of this. Although I do wonder if Open AI is waiting for the most popular applications before trying it for themselves.

What do you think - is originality overrated in business?

Need brutally honest advice on your landing page?

I re-prompted my landing page critique.

This time, I asked it to be sarcastic and harsh…

And combined it with a sarcastic AI-generated voice.

Now, after you enter your landing page, it writes a harsh critique of your landing page.

And it is brutal.

Click the link below ⬇️

(But don’t say I didn’t warn you)

AI is opening the door to a four-day work week
That’s most importantly of all, sustainable.

There’s been a lot of talk about whether a four day working week is really a viable option. Many people are concerned that eliminating the fifth working day would increase workload and stress for the remaining four, but thanks to AI, you theoretically CAN have the best of both worlds.

I’m talking a four day work week, with no changes in workload or productivity.
In fact, it might even make things easier…

Research by the thinktank Autonomy found that:

🔹 Projected productivity gains from the introduction of AI could reduce the working week from 40 to 32 hours
🔹 Pay and performance would remain the same
🔹 Millions of workers (88% of the British workforce) would benefit
🔹 The areas said to benefit most are London, Elmbridge and Wokingham as they apparently hold the highest potential for workers to reduce their hours in the coming decade.

What does this all mean?

AI isn’t a tool to be feared in the workplace ❌

It might actually spark a huge shift towards a better work-life balance - and I think that’s long overdue.

The only downside is that this study projects the changes to not take effect until 2033, but with how quickly the world of AI is advancing, I wouldn’t be surprised if we saw the impacts much sooner. That is, however, only if employers agree. Call me cynical but I doubt many companies would be eager to jump on board with this idea...

What do you think?

(I’ve included a link in the comments to the original study where you can read more!)

Amazon are upping their investment in AI 📈
With a new program called “AI Ready”

It aims to train 2 million workers in AI by 2025, from foundational to advanced level.

They’re also offering a scholarship program for over 50,000 students to increase their AI proficiency.

Amazon said the need for an AI-savvy workforce has never been greater.

And they’re right.

→ A recent AWS study found that hiring people with AI skills was a priority for 73% of businesses
…And yet three out of four businesses can’t meet their talent needs.

The battle for AI-skilled workers has never been greater (why else do you think Open AI pays its researchers over a million dollars?)

So now’s the time to really invest in AI, whether as a business or an individual.

#ai #artificialintelligence #amazon #tech

Edit: Within minutes of posting the below, it's been announced that Altman IS currently in talks to return as CEO of OpenAI.

It's never a dull moment in tech, it'll be interesting to see if this actually goes through or if we'll return to a situation like that below...

In (somewhat) shocking news this week, Sam Altman won’t be returning to his role as CEO of OpenAI.

…And that’s mostly down to Microsoft.

Altman’s firing from OpenAI highlighted a big rift in the artificial intelligence community, with some, like Altman, believing AI is the biggest thing since web browsers, and others worried that we’re moving too fast.

But Satya Nadella’s decision to hire Altman "to lead a new advanced AI research team” at Microsoft shows that those in the ‘pro AI’ camp are strongly winning.

Personally, I think what’s interesting here is that for better or worse, Microsoft is going to get even more control of Open AI. This goes against Sam’s previous statement from a few weeks ago that the board of Open AI should probably be democratized, and somehow be managed by the world, since it would be so powerful.

Yet that power is now potentially in the hands of one of the biggest for-profit companies in the world: Microsoft.

It’s no wonder they raced hard to sign Altman a new deal.


Let me know in the comments below ⬇️

#ai #artificialintelligence #microsoft #openai #samaltman