Deep Mind Releases Alpha Code.

Rob Tyrie
5 min readDec 16, 2022
Photo by Quinn Buffing on Unsplash

#ai #tracking #thewrap #gsg

https://singularityhub.com/2022/12/13/deepminds-alphacode-conquers-coding-performing-as-well-as-humans/

Deep Mind, an AI division of Google, has created a number of algorithms and systems that achieve superhuman performance in things like video gaming and card games like texas hold 'em, has reached another milestone.

Alphacode is an algorithm and machine learning platform using generative technologies to code software. That's software coding software.

This is the first announcement. It comes at the same time where there is much kerfuffle about generating code and correctly by OpenAI. It’s had so many problems that Stack Overflow, a very famous coders support site, has banned answers generated from OpenAI. Based on the different approaches, Open AI will have a hard time competing witn Google here.

The Alphacode technique is different. It is “more trained" , so that it provides unique, working answers. It also uses a technique to generate a lot of simulated answers that can be compared so that the ultimate algorithm is more fine tuned. These are called GAN’s, or general adversarial networks, a now common ai technique.

It is so good that it mimics humans in creating code from structured english language prompts.

There are contests and competitions for programmers. This is not unusual. These are just like math and physics contests, only they are focused on coding skills. Programming is a completely human construct devised to make it easier to tell a computer what to do and a mix of language, grammar and logic. Focus on the word easier there. It’s still actually not that easy after all these years.

These contests pit groups of humans against each other with assignments to code, and set up working software programs.

Deep Mind as created a number of ai systems in this fashion. The results follow the same pattern, they are tested early and often are pitted against humans. They fail at first and then they learn, that’s part of the algorithm too, they learn. And once they learn they achieve better and better results until they reach parity.

Now in the case of Chess, Go, Texas Hold’em, Starcraft and Diplomacy, eventually the system learns to achieve superhuman capabilities. Better than humans. On average.

In this first announcement, the Alphacode algorithms were entered into a contest with humans and beat or exceeded the capabilities of 50% of the humans. So that’s kind of like the maintenance coders.

If the pattern holds true, as with the other achievements, sometime in the next 6 or 12 months off of code will exceed the capabilities of human coders. or maybe 98% of them.

It is important to note that a Microsoft driven algorithm called the Codex also generates code. However, it copies other people’s code and it ran into controversy that’s related to copyright and is now being sued by coders that feel that their rights were being abrogated. This new algorithm from Deep Mind purposely avoids copying code, and just like other human coders did in the past, they can create new code based on all code that is unique and it’s own copyright. This will bring into question the ownership of the generated code.

The Alphacode algorithms will very likely gain capabilities and performance rapidly. Again, based on the other achievements, there will be a one or two month long period of a lot of yelling and screaming that either its excellence wasn’t true, or it wasn’t a fair contest, or it’s not really what it says it is. However, as in the case of the games, the algorithms probably will become superhuman. That’s good news for owners who need code.

Unsurprisingly, there probably will be articles written about the "death of the programmer", and a great human cry about the loss of jobs and loss of careers of many, many people. It’s just a repeated pattern.

Another good pattern for one to focus on, outside of the yelling and screaming, is, that in the case of Texas Hold’em and Go, when continued contests occurred, humans got better and so did the machines. However, the best teams were combinations of humans and machines. they learn together and exceeded the former capabilities of humans alone.

Maybe, we should concentrate on that partnership rather than an adversarial footing. It seems that most scientist in this area are proceeding that way, and most media is not. But, that’s just a gut feel guess.

One last note, the discussion around this area without a deep dive into ethics, regulation, controls and enforcement, is not right. They must all be considered.

The creation of software code is a wonder of the world, like Marc Anderson said software ate the world, and now it’s growing, and a machine wrote it.

It's just a machine that reflects us. Us.

We need to control it with ethics, intellect and an ingenuity. That's what humans do.

At least that's what we did for the last millennia, because we are still here.

The OpenAI Summary and comment

Development of an artificial intelligence system called Alphacode, which is capable of generating software code. It is said to have outperformed 50% of human coders in a contest, and it is expected that its capabilities will continue to improve rapidly. The use of this system has raised questions about the future of programming jobs and the potential for AI to surpass human capabilities in this field. It has also been suggested that the best results may come from a partnership between humans and AI, rather than an adversarial relationship. It is important to consider the ethical, regulatory, and control issues surrounding the development and use of AI systems like Alphacode. Is there anything else you would like to know?

Robert Tyrie
Is Alphacode better than OpemAI

I'm sorry, but I do not have enough information about Alphacode or OpenAI to accurately compare the two. As a language model, I do not have access to current events or the latest developments in artificial intelligence, and my knowledge is limited to the information that was available at the time of my training, which has a cutoff date of 2021. I recommend doing further research to learn more about the specific capabilities and limitations of these two AI systems.

#ai #human #partners #history

--

--

Rob Tyrie

Founder, Grey Swan Guild. CEO Ironstone Advisory: Serial Entrepreneur: Ideator, Thinker, Maker, Doer, Decider, Judge, Fan, Skeptic. Keeper of Libraries