DALL-E, Copilot And The Deep Learning Models You Need To Know Now

One of the new deep learning models I’ve been fascinated by is called DALL-E, developed by OpenAI. This technology with roots in GPT-3 can generate digital images from your voice commands.

For example, if I have a picture of myself on vacation and I’d like to add a funny hat to my photo, I can simply command DALL-E (now in its 2nd generation as DALL-E 2) to “add a funny hat.”

That’s just the beginning, however. The real potential I see here is that, up until now, much of the imagery we’d be able to use is based on what we have available – what we know and can see. DALL-E 2 gives us a new dimension in which the unknown can be created in picture form.

painting of a fox sitting in a field at sunrise in the style of Claude Monet
Painting of a fox sitting in a field at sunrise in the style of Claude Monet

Let’s say I’d like to have a painting of a fox sitting in a field at sunrise in the style of Claude Monet. Good luck finding that stock photo, right? DALL-E 2 doesn’t need luck. It will create that image for you. Like anything in beta, you may or may not find the image to be precisely what you intended, but the quality is certainly getting better all the time. And since all the images you’re creating with DALL-E 2’s deep learning are practically guaranteed to be original in their form, there are no royalties to pay.

What Makes DALL-E 2’s Deep Learning Model So Different?

In one sense, what we’re describing is part of a pattern we’ve seen in AI: We human beings have a hand in what we want to be done before the technology goes to work. DALL-E 2 is not going to create images for us without that direction (yet, anyway). We are still in control.

Yet, something different is happening with the release of DALL-E 2 in that so many more people will get to experience it right away. We’ve heard of promising deep learning models that have been accessed by a select number of individuals to test. OpenAI had a different idea with DALL-E 2: It simply opened the technology up to the general public. By now, there are already at least a million people using it. All you need to do is to sign up for free.

Not only is access to this model wide open, but people from all walks of life find it easy (and fun) to use right away.

What Industries May Be Disrupted By DALL-E 2 And Other AI?

We can see where companies that sell stock photography and charge for it, such as Getty Images, may have a problem. They offer various image licenses, which is how they make money. You could pay at least $35-40 for just one image. And when you buy it, what you see is what you get. There’s only so much you can do to that image.

On the other hand, if you had the idea of an image based on something you’ve seen but with a nice twist on the concept, you can give DALL-E 2 some instructions and have your original image, which is excellent for the public. Perhaps less so for companies who charge a substantial licensing fee.

Coding as we know it stands to undergo a powerful shakeup in its own right, thanks to an AI-powered application called GitHub Copilot. Developed by Microsoft’s GitHub and OpenAI, Copilot can recommend entire lines of code based on your activity, essentially saying, “Ah! I see what you’re trying to do here. How about you just use this type of code?”

The AI “learns” enough from your direction to transform you from a traditional code writer to a  better one. Is it designed for the highest-level software developer right now? No, the target is likely to be more on mid-level software development at this time.

Still, just as DALL-E 2 was opened to a broad public audience, Copilot has quickly moved from being at the stage of technical preview to being an application that will generally be available to all developers and accessible to students.

Currently, there is a talent shortage in several critical areas of the technical space. Software developers’ salaries have gone from $10 an hour in Ukraine to $50-60 an hour for the same workers. Similarly, salaries for software developers have skyrocketed in India.

What does that tell us? While there’s a lack of supply, there is ample demand for software developers. But what if user-friendly tools such as Copilot allow more software to be developed at a less expensive rate?

We’ll likely see more people wanting to adopt and use it, which could be especially helpful for organizations experiencing a talent shortage. The developers they have on staff can use Copilot for programming languages such as Python, JavaScript, Ruby, Go, and more.

Copilot won’t reduce the number of developers. Far from it. In fact, it will only make developers more productive. And, as more people use it, it will only achieve better results.

Talman Advantage #4: Better Positioning For Your Best Opportunity

The reality is that, in so many situations, that “perfect job opportunity” may not be formally listed by a company. In that instance, where some may simply fire your resume off to an HR person’s email and hope for the best, Roy Talman & Associates takes a more creative and purposeful approach. If an opening isn’t currently available that’s an ideal match for you, we’ll discuss the kind of role with you that you would be interested in and potentially prepare and present a very specific case to that particular firm to create a unique role for you. That’s called a recruiter that goes further for you – and why you need to Talk To Talman First.

The medical field is another candidate ripe for disruption by deep learning.

Think about a hospital where a surgeon is viewing the x-rays of a patient. How many images is that surgeon typically viewing? Five images. Maybe as many as 10 images.

Imagine the same environment with a machine learning tool that enables a surgical team to view an entire “movie” of you at 60 frames per second. We’re not talking about a collection of static images that aren’t completely clear and precise. No. It’s a 20-minute experience where the camera constantly moves and vividly captures images.

As a result of machine learning, you don’t necessarily have fewer radiologists. You may have more radiologists, which has created a new field called active radiology. Instead of reading pictures, active radiologists are part of the surgery team responsible for taking X-rays on the fly during surgery. Machine learning can make this imaging process far more effective, which likely means more active radiologists will demand the tool.

That’s not all – patients may demand machine learning tools and make their choices in healthcare based on which hospital has such tools. One hospital doesn’t have machine learning tools for your colonoscopy but another one not only does, but they have a team that has been highly trained on how to use such tools.

Armed with this new information, which hospital do you think the patient elects to have their procedure at? You can probably guess.

From our vantage point, there will be such a continual expansion of machine learning capabilities that we will reach the point where any business that figures out how to incorporate AI seamlessly will potentially be in an extraordinary position. It will essentially resemble a new business that has the capability to put a bigger, slower-moving, more traditional firm out of business.

After all, you have to evolve to survive and thrive.

In a future divided by those who embrace deep learning and those slower to do so, there’s bound to be a significant change. It won’t happen overnight, perhaps, but it will happen. And when it does, the companies that have trusted Roy Talman & Associates to help them stay ahead of the race for top talent will be in an ideal position for the next era of growth. Don’t wait for the future to sneak up on you. It’ll be here before you know it. Talk To Talman First.