Category Archives: Commentary

An Open-Source CUDA for AMD GPUs – ZLUDA

Lots of work has been put into making AMD designed GPUs to work nicely with GPU accelerated frameworks like PyTorch. Despite this, getting performant code on non-NVIDIA graphics cards can be challenging for both users and developers. Even in the case where the developer has appropriately optimised for each platform there are often gaps in performance where, at the driver-level, instructions to the GPU may not be optimised fully. This is because software developed using CUDA can benefit from optimisations like operation-fusing without having to specify in many cases.

This may not be much of a concern for most researchers as we simply use what is available to us. Most of the time this is usually NVIDIA GPUs and there is hardly a choice to it. NVIDIA is aware of this and prices their products accordingly. Part of the problem is that system designers just dont have an incentive to build AMD platfroms other than for highly specialised machines.

Continue reading

Fail fast

While scrolling through my Instagram reels feed, I came across a reel of Jensen Huang, NVIDIA’s CEO, talking about the need to fail fast, which motivated me to write a post. ‘Fail fast’ is a recent piece of advice I have been hearing since I embarked on my PhD; fail fast on the research directions that we plan to pursue so that we can understand the difficulties and limitations of the research problems and methods used which will in turn give us more time to finetune our problem and develop more nuanced approaches. Since childhood, most of us have been taught that failures eventually lead to success and that persevering towards success is critical. However, one thing that I could not come to terms with is the narrative of several failures ‘magically’ leading to success. If you were destined to be successful, why would you even fail? And also, for every failure-to-success story we hear, there are many other stories of failure that we don’t.

Continue reading

250 Trips on the Oxford Tube: what I’ve learnt

The Oxford Tube is a bus service that shuttles people between Oxford and London taking approximately 1 hour and 30 minutes. I have now taken the bus over 250 times which is approximately 375 hours or a fortnight of my life.

In this time spent on the bus, I have discovered some tips and tricks that make the journey ever so slightly more bearable. I shall share them so that others can optimise their experience. Enjoy!

Continue reading

On National AI strategies


Recently, I have become quite interested in how countries have been shaping their national AI strategies or frameworks. Since the launch of ChatGPT, several concerns have been raised about AI safety and how such groundbreaking AI technologies could augment or adversely affect our daily lives. To address the public’s concerns and set standards and practices for AI development, some countries have recently released their national AI frameworks. As a budding academic researcher in this space who is keen to make AI more useful for medicine and healthcare, there are two key aspects from the few frameworks I have looked at (specifically the US, UK and Singapore) that are of interest to me, namely, the multi-stakeholder approach and focus on AI education which I will delve further into in this post.

Continue reading

OPIG: A decade of Scientific Shenanigans. What’s changed?

2013 was a big year: Andy Murray clinched the Wimbledon title, NASA’s Curiosity Rover discovered water-bearing minerals on Mars, and ‘twerk’ and ‘selfie’ made their way into the dictionary, something equally significant happened—the birth of BLOPIG.com. Intrigued by how the group has changed over the last decade, I started on a journey to unearth the some of the publications from then till now. Questioning their focus, methods, and evolution of the groups’ research over the past decade. This blog post is what I found.

While delving into each publication of the past decade genuinely seemed like an interesting idea, the imminent threat to my PhD progress forced me to adopt the most 2023-appropriate approach: outsourcing the task to AI. After collecting abstracts from all the group’s papers, I enlisted the help of everyone’s’ favourite hallucinator to summarise the works and (hopefully) highlight the shifts in their research.

So after a relatively long, sequence of prompts, this is (apparently) what we do?

Continue reading

Lucubration or Gaslighting?​

Or: The best lies have a nugget of truth in them.​

Lucubration – The action or occupation of intensive study originally by candle or lamplight.

Gaslighting – Psychological abuse in which a person or group causes someone to question their own sanity, memories, or perception.

I was recently having a play with Google Bard. Bard, unlike ChatGPT has access to live data. It also undergoes live feedback and quality control. I was hoping to see if it would find me any journals with articles on prion research which I’d previously overlooked.

Me: Please show me some recent articles about prion research.
(Because always be polite to our AI overlords, they’ll remember!)

Continue reading

“The Rise of ChatGPT 4.0: Is the Future of Work in Jeopardy?”

In my previous blog post, I explored the capabilities of ChatGPT 3.5, testing its skills as a programmer and mathematician’s assistant. The results were mixed, to say the least. While it could handle simple coding tasks with ease, it faltered when faced with more complex mathematical problems and image manipulation tasks. I concluded that while ChatGPT 3.5 was impressive, it was far from replacing us in our jobs. It was a tool, yes, but one that needed careful supervision and frequent corrections.

Fast forward to today, and we have ChatGPT 4.0 with code interpreter enabled. I approached this new version with the same scepticism I had for its predecessor, but I must admit, my initial impressions have been challenged. The leap from 3.5 to 4.0 is not just a step forward; it’s a giant leap. You can see my experiment here https://chat.openai.com/share/0099a12f-e577-44bd-aa10-7c7508f15f19.

In my previous experiment, I asked ChatGPT 3.5 to translate a 2D image, a task that combines both programming and a bit of maths. The results were less than satisfactory. It was able to write code to translate images if they were in real space, a task that can be easily found searching on the web. However, it struggled to handle images represented in Fourier space correctly, and even with multiple hints, the best it could do was a rough approximation of the correct solution. It took six attempts to get the code automatically generated, and even then, it was far from perfect.

With ChatGPT 4.0, however, the story is different. When asked to perform a similar task, it provided a clear, concise, and correct solution. It understood the concept of image translation in the context of computer vision and image processing, and it was able to generate Python code using the scikit-image package to perform the task. The code was not only syntactically correct but also logically sound, demonstrating a significant improvement in understanding and problem-solving capabilities.

"""
#Me: Can you know provide some python code to translate (pixel shifts) images? And can you execute it?
#Chat gpt: Here's a simple Python example using scikit-image to translate an image:
"""

import numpy as np
import matplotlib.pyplot as plt
from skimage import data
from skimage.transform import AffineTransform, warp

# Load an example image
image = data.camera()

# Define a translation transform
tform = AffineTransform(translation=(50, 50))

# Apply the transform to the image
translated_image = warp(image, tform.inverse)

# Plot the original and translated images
fig, ax = plt.subplots(1, 2)
ax[0].imshow(image, cmap='gray')
ax[0].set_title('Original Image')
ax[1].imshow(translated_image, cmap='gray')
ax[1].set_title('Translated Image')
plt.show()

"""
#Chat gpt: 
In this code, AffineTransform is used to define a translation of 50 pixels to the right and 50 pixels down, and warp is used to apply this transform to the image. The result is an image that has been shifted.
However, I must clarify that the current environment doesn't support displaying plots, so the matplotlib part of the code won't work here. You should be able to run it in a Jupyter notebook or similar environment on your local machine. Let me know if you need further help or clarification!
"""


"""
ME: perform the same operation as before but with an image represented in Fourier space. 
#Chat gpt: 
Sure! However, we need to keep in mind that translating an image in the Fourier domain is not the same as translating an image in the spatial domain. In the Fourier domain, translations are performed by multiplying the image by a complex exponential, which essentially changes the phase of the Fourier transform.

Here is a Python code snippet demonstrating translation in the Fourier domain:


"""

import numpy as np
from skimage import data
from scipy.fft import fft2, ifft2, fftshift, ifftshift

# Load an example image
image = data.camera()

# Compute the 2-dimensional FFT of the image
image_fft = fft2(image)

# Define a translation
dy, dx = 30, 30

# Generate an array with the same size as the image
x = np.arange(0, image.shape[1])
y = np.arange(0, image.shape[0])
xv, yv = np.meshgrid(x, y)

# Create the translation function in Fourier space
translation = np.exp(-2j * np.pi * (dx * xv / image.shape[1] + dy * yv / image.shape[0]))

# Translate the image in Fourier space
translated_image_fft = image_fft * translation

# Transform back to spatial domain
translated_image = np.abs(ifft2(translated_image_fft))

# Now you could plot the images
"""
#Chat gpt: 
This code first computes the Fourier transform of the image using fft2. Then it generates a complex exponential function that represents a shift in the Fourier domain. This function is multiplied with the Fourier-transformed image to achieve the translation. Finally, the inverse Fourier transform ifft2 is used to transform the image back to the spatial domain.

As before, the current environment doesn't support displaying plots, so you would need to run this on your local machine to see the translated image.
"""

Another test I revisited was a mathematical problem I had posed to ChatGPT 3.5: “What is the intersection between the unit circle and the curve y = x^2?” The previous version had failed to provide a correct answer. However, ChatGPT 4.0 not only understood the problem but also wrote its own code to solve it, correctly identifying the intersection points. This was a clear demonstration of the AI’s improved mathematical problem-solving capabilities although It’s important to note that only the premium version of ChatGPT 4.0 can run the code it suggests, as it did in the example of the intersection of the unit circle and the parabola. +

  • ME: Can you tell me which is the intersection between the unit circle and the curve “y= x*x”?
  • ChatGPT: The unit circle is defined by the equation x2+y2=x2+y2=1. The curve y=x2 is a parabola. To find their intersection, we can substitute y2 from the equation of the parabola into the equation of the circle, and solve for x2. So, we have x2+(x²)²= 1 or x²+x⁴=1 This is a quartic equation and can be difficult to solve, but let’s give it a try. [AND IT STARTS WRITING AND EXECUTING CODE]

So, is my job in danger now? Well, it’s a complex question. The advancements in AI, as demonstrated by the jump from ChatGPT 3.5 to 4.0, are indeed impressive. The AI’s ability to understand complex tasks and generate accurate solutions is growing quite fast. However, it’s important to remember that AI, at its core, is a tool. It’s a tool that can augment our capabilities, automate mundane tasks, and help us solve complex problems. In the end, whether AI becomes a threat or an ally in our jobs depends largely on how we choose to use it. If we see it as a tool to enhance our skills and productivity, then there’s no danger, only opportunity. But if we see it as a replacement for human intelligence and creativity, then we might indeed have cause for concern. For now, though, I believe we’re safe. The Turing test might be a thing of the past, but the “human test” is still very much alive.

How ChatGPT changed my writing as an ESL speaker

It’s not always easy to live in an Anglophone scientific world when English isn’t your first language. When careers are built upon the ability to communicate ideas clearly and eloquently, struggling to find the right words can be a real hindrance to explain your science in a way that is taken seriously. Contrary to popular belief, it’s not something you can simply “work” on. Often, it doesn’t matter how many books you’ve read, how many years of education you have, or how articulate you are in your original language — your brain will refuse to summon the right expression, or get stuck in a construction that a native speaker would never use. Struggling with a second language is very much a biological phenomenon.

The standard recommendation for ESL (English as a Second Language) speakers has long been to ask a native colleague to read through any text that needs to be published or submitted somewhere (such as an article or a grant application). Well-intentioned as this advice may be, there are multiple problems with it. Lingua franca or not, only 15% of the world population speaks English, of which only 5% are native speakers — meaning that for most scientists not working in Anglophone countries, the option is rarely available. Even when available, it is unreasonable to expect these colleagues to add charitable proof-reading to their workload simply because they happened to be born speaking a different language. But, most importantly, I have always felt — and I want to emphasize that I truly believe most people who issue this kind of advice to be well-intentioned — that the underliying message sounds too much like “you need vetting by a member of our select linguistic club if you want your ideas to be taken seriously“.

Continue reading

Happy 10th Birthday, Blopig!

OPIG recently celebrated its 20th year; and on 10 January 2023 I gave a talk just a day before the 10th anniversary of BLOPIG’s first blog post. It’s worth reflecting on what’s stayed the same and what’s changed since then.

Continue reading