Convergence of talent and AI tech will make India a powerhouse of AI, says GitHub CEO, Thomas Dohmke


Microsoft-backed developer platform GitHub has over 15.4 million developers in India, growing 33 per cent year-over-year (y-o-y). India is set to pass the US as the largest developer community on GitHub by 2027. In an interview with businessline, GitHub CEO Thomas Dohmke highlighted that India’s burgeoning developer community, combined with the newfound possibilities of AI, will not only accelerate digital transformation, but will drive immense human and economic progress for India. 

With AI coming in big into coding and software development, what does it mean to be a software engineer in this day and age compared to, say, the dot com era ?

Being a software developer today means that you have to manage incredibly complex systems. You can’t imagine both your personal and professional life without software anymore. Every company today is now a software company. So developers are sitting on this ever-growing pile of software and systems, and the only way to deal with this complexity is to go to the next level. We believe AI will allow us to do so in the same way that developers have moved from punch cards to assembly language to basic programming languages like Basic and Pascal, and then to higher programming languages, to the strong adoption of open source and the adoption of the cloud. With our AI co-pilot, we are seeing productivity gains upto 55 per cent.

But if AI is already doing coding, what does the future hold for developers?

We’re not yet at the stage where AI can write all the code. While it can generate some lines, developers must take these ‘Lego blocks’ produced by AI and assemble them into a cohesive set. This process remains an art in itself. Most software doesn’t adhere to instruction booklets; it stems from the creative minds of developers, program managers, designers and others. One way to conceptualise this is through systems thinking. Developers must tackle complex issues, break them down and solve smaller problems, deciding when to leverage AI. As AI capabilities expand, it’s akin to having larger Lego blocks available. However, developers must still make decisions — like choosing a database or cloud technology, or determining whether the software should run on a mobile phone, a point-of-sale system, or a car. These decisions require human creativity, intuition and judgement. Presently, AI models lack these features and only time will reveal if we’ll ever reach that stage. Nonetheless, it’s currently inconceivable for developers to be replaced by AI. AI serves as a companion, a co-pilot, assisting with tasks that developers prefer not to handle, enabling them to focus on the creative aspects that bring value to companies.

Will AI democratise coding that will enable billions to engage with technology as easily as driving a bicycle? What does this mean for a country like India?

 I think it will impact the economy the same way the Industrial Revolution did 200 years ago. It will completely shift the way we’re thinking about productivity in our companies, and in our lives.

In India, you have a great convergence — on the one side, there’s this new technology that is in its early stages and on the other side there is a big number of computer science graduates every year. In the next few years, these two things will make India a powerhouse of AI.  What India needs is for its kids to learn coding as they learn art, science and math. There is no future where we don’t have computers around us everywhere.

Second aspect is to have clear regulation and policies, especially around large language models (LLMs). Regulation should let open source maintainers, researchers, students and teachers to do to do their work without the same compliance requirements as big companies. If I’m an individual in Bengaluru working on an open-source project, I cannot afford to also have a compliance department.

Third, companies need to invest in AI.  The time is now, the train is already leaving the station.

The developer ecosystem in India on GitHub has crossed the 15 million mark. What is driving this growth?

The accessibility of technology has become so much easier. Today, everybody has a cell phone that is connected to the Internet. We have an interconnected community of developers all around the world that work with each other, without any borders and without any boundaries. All that matters is the code and the intent that you have to contribute back. In India, people see this as an opportunity to improve their lives. It allows them to build a reputation for being an open-source software developer.

How do you see the adoption of AI by Indian companies?

We are really happy with how the AI adoption is going on and also about the opportunity that lies ahead for India to leverage. That’s part of the reason why I’m here… because we see tremendous customer interest.

How do you see the ongoing geopolitical situations such as sanctions against some countries, or the fact that China has developed its own rival to GitHub?

Open source is the only community where politics play almost no role. We are fighting for the rights of of developers everywhere

What keeps you awake and what excites you the most?

Believe it or not, some nights I’m awake because I’m so excited about the opportunity ahead. I’ve been a developer since I grew up in East Germany. I couldn’t just go and buy a computer. I had access to an East German computer in the geography lab in school. So once a week I was allowed to code a little bit. So I’m excited about the opportunity that everybody today has if they want to become a software developer. Nothing is stopping them. The Internet, the cloud, has enabled a world of collaboration, sharing and creation. Now AI is transforming this world and it’s giving everybody the same opportunity. The flip side of that is that as the systems become more complex, we also have to constantly think about securing the software supply chain which cannot be exposed by bad actors.

Published on June 11, 2024





Source link

The secret intelligence of plants


Ever since Peter Tompkins and Christopher Bird wrote in their book The Secret Life of Plants, about an experiment of a former CIA polygraph expert, Cleve Backster, things have never been the same in people’s thinking about plant consciousness and intelligence.

Backster hooked a polygraph (lie detector) machine to a plant, sat near it and began thinking evil thoughts, like setting fire to the plant. The polygraph “went crazy”; Backster reasoned that the plant could read his mind.

Later, when others tried the same experiment and failed, he is said to have observed that you first need to develop a rapport with the plant.

Regardless, after The Secret Life of Plants, many have developed the conviction that plants could think. Some play classical music to them to make them yield better fruits.

Zoe Schlanger — a climate report at The Atlantic and author of The Light Eaters, which explores plant intelligence — has described in an interview about an extremely interesting experiment made by Heidi Appel of the University of Ohio. Knowing that plants respond to caterpillars chewing on them by sampling the caterpillar saliva and using the information to generate the exact chemical compounds that attract parasitic wasps that kill the caterpillar by injecting their eggs into it, Heidi recorded caterpillar chewing sounds and played them to the plants. The plant responded exactly as how they would if they were being chewed by real caterpillars.

A person known to this writer once teased a young creeper by moving the support all around (over many days). After a few attempts the creeper stopped following the support, leading him to conclude that creepers have “self-respect”.

Well, the evidence on plant intelligence is far from conclusive, but perhaps enough for us to be kind to plants.





Source link

Next-generation anodes fuelling electric dreams


How nice it would be if one could charge the battery of an electric car and drive without a care for 1,000 km!

A battery so dense with energy that can last really long is at the heart of electric mobility. To get there, it is necessary to have energy-packed electrochemical cells (a battery is made of cells.)

When you charge a cell, the electrons from the cathode split from their parent atoms; the cathode then is left with positive-charged ions. The negatively charged electrons run along an external circuit and reach the anode; meanwhile, the ions run through the electrolyte and also reach the anode.

The crux of making a battery that lasts for 500 km is in making an anode that can accommodate a heck of a lot of ions. The more room it has for the ions, the more energy it can pack.

Today, carbon (in the form of graphite) is the anode material of choice. For cathode, the common choices are different compounds of lithium. The lithium is ‘housed’ in other materials; when its ions cross over, they again get housed in the anode, which is almost all the time. Now, the search is for anode materials that have more room to seat lithium ions. There are many known alternatives for graphite, but each comes with its own flipside — which is why graphite is still the king. But some nice flavours are emanating from research labs in the recent times. Businesses have begun to take these new anodes seriously enough to work out business models and some investments are also beginning to happen. The lithium-ion batteries that use these new anodes are called ‘Advanced Li-batteries. In this article, we will take a look at two of them.

Our old friend, Silicon

This ubiquitous stuff from sand, a big friend of mankind, in use everywhere from glass to semiconductors, is a darn good anode material. Pure silicon can accommodate ten times as many ions as graphite does. Unlike with graphite, where lithium ions embed themselves in its ‘pores’ (called ‘intercalation’), the ions form an alloy with silicon (called ‘conversion’) — where the energy is stored in the bonds between lithium and silicon.

This use of silicon has been known to scientists for over six decades. The industry has been sniffing at silicon to make anodes for about twenty years. Many companies have considered it and given it up.

Why? Because silicon has a few disadvantages. As it takes in lithium-ions while charging, it expands. While discharging, it contracts. This causes pulverisation of large silicon particles, which affects the integrity of the anode. Second, silicon has low electrical conductivity (ions do not swim through it easily, unlike in carbon). Third, it affects the Solid Electrolyte Interphase (SEI) — a protective film that quickly forms on the surface of the anode that touches the electrolyte and then permits only ions and shuts out any wandering electrons.

Due to these issues, some companies just sighed and walked away (towards the other promising anode that we will discuss later in this article.)

In a 2021 whitepaper titled ‘National Blueprint for Lithium Batteries’, the Federal Consortium for Advanced Batteries, US, does not mention the word ‘silicon’ even once. A point to note is that even batteries of today have a little bit of silicon in their anodes, to increase energy density.

The solution

Some companies, however, have been doggedly pursuing silicon to make anodes for several years now. “Many companies have tried to address the limitations of silicon as an anode in different ways,” notes Dr Rahul Gopalakrishnan, CEO at ABEE group of Belgium, “but still these are at lab level or pre-manufacturing stage, somewhere at TRL 4-6.” The Abee group is into the other type of anodes — lithium metal.

He notes that the trick is in scaling up. Leaping from lab to industry is fraught with challenges, he says. “If this bridge is crossed we will see widespread adoption of silicon as a viable alternative to graphite as an anode,” Gopalakrishnan tells quantum.

But it appears that some companies have cracked the code of scaling up.

Sicona, an Australian start-up, is building a silicon carbon anode plant in the US. Last year, the Kolkata-headquartered Himadri Speciality took a 12.8 per cent stake in Sicona.

Sicona has not given out much about its technology, but it is known that the core of its technology is based on silicon nanoparticles and a “specialised coating process to create unique material qualities.” Nano particles are so small that while they may expand and contract, they won’t pulverize — you can’t pulverize powder. Also, clearly, Sicon’s anode is not pure silicon, as it says it will deliver a 20 per cent increase in energy density.

Another notable company is the California based Sila Nanotechnologies, founded by Gene Berdichevsky, an ex-Tesla employee. Bessemer Venture and Canada Pension backed Sila describes its technology thus: “What’s required to replace graphite entirely is a material that compensates for the swelling of silicon through the design of an engineered particle structure. If you can create a particle that allows the swelling and contraction of silicon to happen inside the particle, while keeping the electrolyte outside of the particle, you could cycle the material reversibly 1,000’s of times and perhaps 10,000 times.”

Sila is already in production; in December 2023, it signed a commercial agreement with Panasonic to supply silicon anode materials. (Quantity and price have not been disclosed.)

Yet another company into silicon anode is Amprius Technologies, USA, which also speaks of silicon nanowires and special coating. Amprius says that its battery of energy density of 370 Wh/kg of silicon anode material is already commercially available; the battery is capable of “extreme fast charge”— up to 80 per cent within six minutes.

So, it does appear that silicon has been tuned for batteries.

Lithium metal

The companies that have said “to hell with silicon” have turned to the other promising anode — lithium itself. By having a copper foil on the anode end and charging, the lithium ions — instead of intercalating (graphite) or forming bonds (silicon) — just plates itself on the foil.

The beauty of a lithium metal anode (or, anode-less batteries) is its high energy density — 4,600 Wh/kg. But, as with silicon, there are issues — mainly formation of dendrites (spikes) that can pierce the electrolyte and touch the cathode. “The large scale of lithium metal also faces engineering problems such as mechanical stress by coiling and slitting, electrolyte consumption and volume swelling stress,” says a recent scientific paper. These issues are not insurmountable but they call for a new supply chain. “The commercial use of lithium metal batteries has a long way to run,” the paper notes.

Yet, as in the case of silicon, the challenges have not totally deterred businesses from trying their hand at lithium metal anodes. ABEE is an example.

Beyond lithium

Is the world resigned to having (the scarce) lithium for cathode? Not quite. Today, lithium is the best — challengers such as sodium, aluminium and zinc can at best be niche players. But you can’t rule out other equal matches for lithium. Scientists are thinking in terms of metal fluorides (iron fluoride, copper fluoride) and sulphur-based cathodes.

Research is also on to make solid electrolytes to replace the incumbent liquid — an area of research that has both strong supporters and sceptics.

The broad message is clear. In the not-too-distant future, you would be able to travel 1,000 km on one charge.





Source link

Agnibaan, the rocket that defied the odds


First-time-right is not common with anything about space, certainly not for a newly-designed rocket, with the engine minted on a 3D printer. But if the Chennai-based, IIT Madras incubated start-up, Agnikul Cosmos, got the maiden flight of its Agnibaan rocket right, it was because of the mountain of preparation that preceded the launch.

“We did 20,000 computer simulations to study what could go wrong with the trajectory,” says a delighted Dr Satyanarayanan Chakravarthy, professor of aerospace and combustion engineering at IIT Madras, who mentored Agnikul to the successful launch on the morning of June 1. Each simulation checked the behaviour of the rocket against a certain variable, such as a gust of wind from a certain direction.

Alongside, they did about 40 static tests to see how the machine works under hot-fire conditions — especially the gimbal mechanism that ensures the stability of the rocket. In the end, the flight that took the rocket to a height of 6 km and let it fall into the sea 2 km from the launch pad was a success. “Very few rockets have had success in its maiden flight,” Chakravarthy told quantum.

True, Agnikul launch was aborted thrice — twice shortly before the planned lift-off, but those aborted launches only demonstrated that the Automated Launch System was robust and could catch any bugs.

The rocking rocket

A few points about the rocket must be kept in mind.

First, it was India’s first rocket that was powered by liquid fuel in the core. All the ISRO rockets have solid fuel in the core, though the strap-on rockets (those little ones that cling to the sides of the main rocket at the bottom) were liquid-fuel fired. Incidentally, the Agnibaan is also designed to be fitted with strap-ons, when heavy payloads ask for it. Second, it was for the first time in India that a semi-cryogenic engine was used — ATF (fuel) at ambient temperature and liquid oxygen (oxidiser) in cryogenic condition. This meant that the fuel loading had to begin only 3 hours before the lift-off.

Third, what flew on June 1 was not the full 2-stage rocket, but the top half of it with a single engine. The full rocket, with a cluster of 4-7 engines in the lower stage, would be test-launched later. Fourth, it was also the first private rocket that was controlled during the entire flight — its velocity, attitude and position fully telemetered, and could be destroyed by a person on the ground if it went awry. The other rocket start-up, Skyroot Aerospace, test-flew its rocket in November 2022, but it was a sounding rocket that just went up and came down. (This is not to say that Skyroot’s vehicle is inferior to Agnikul’s or Skyroot lags in technology—different companies adopt different pathways for developing a vehicle.)

Extreme caution marked the making of the rocket. First, it was powered for a thrust of 1.1, which meant that the thrust was just a little more than its weight — as such, the rocket didn’t shoot-off into the skies but ascended slowly. More thrust would have meant burning more fuel at the launch pad. Also, for design simplicity, it was an un-throttled engine — no throttle to adjust the thrust by controlling the flow of the fuel and the oxidiser into the combustion chamber.

Second, the launch sequence was so designed that ignition would happen 7 seconds before lift-off and while it would take only two seconds for enough pressure to build up inside the combustion chamber, the vehicle waited for a full five seconds before leaving the ground.

The next steps for Agnikul is to build the full, two-stage vehicle and demonstrate stage-separation, on the ground. Chakravarthy believes the company would be able to come to this point in nine months.





Source link

How chatbots help avoid embarrassment online


Imagine shopping online for a personal item you’d rather keep private. Would you prefer talking to a human or a chatbot?

Recent researches show that chatbots — when clearly identified and not overly humanised — help consumers feel less embarrassed during such purchases. The paper “Avoiding Embarrassment Online: Response to and Inferences about Chatbots when Purchases Activate Self-Presentation Concerns” by Jianna Jin, Jesse Walker, and Rebecca Walker Reczek, published in the Journal of Consumer Psychology investigates this question.

Role of ambiguity

The rise of artificial intelligence (AI) in customer service has revolutionised how people shop online. Chatbots, a type of conversational AI, often operate behind the scenes, sometimes with ambiguous identities that can lead consumers to believe that they are interacting with a human. This ambiguity becomes particularly relevant when self-presentation concerns are at play.

Self-presentation is about controlling the impressions others form of us, especially during potentially embarrassing purchases. Jianna Jin and her colleagues explored how these concerns affect consumer interactions with chatbots versus human agents. They hypothesised that consumers with high self-presentation concerns would see an ambiguous chat agent to be human to brace for potential embarrassment.

The study confirmed this hypothesis: consumers with higher self-presentation concerns were more likely to assume an ambiguous agent was human. This inference serves as a psychological buffer, allowing them to prepare for any potential judgment, even if the agent turns out to be a bot. This aligns with the principles of Error Management Theory, which suggests that people make biased inferences under uncertainty to avoid more costly errors.

Comfort of Knowing

The research also examined consumer responses to clearly identified chatbots. Contrary to earlier findings suggesting negative reactions to known chatbots, Jin and her team found that consumers preferred clearly identified chatbots over human agents when self-presentation concerns were active. This preference stems from the perception that chatbots possess less “mind”— less capacity for consciousness and emotional judgment — compared to humans.

When consumers know they are interacting with a chatbot, they feel less embarrassed because they believe chatbots lack the emotional and cognitive depth to judge them. However, the study reveals a crucial nuance: the design of the chatbot matters. Anthropomorphised chatbots, which exhibit human-like qualities, increase consumers’ feelings of embarrassment. Thus, less human-like chatbots are better suited for sensitive purchases.

The Perfect Chatbot

The findings have significant implications for businesses aiming to improve customer experience, particularly in sensitive contexts. Clearly identifying chatbots can reduce consumer embarrassment and facilitate interactions that might otherwise be avoided.

Businesses should consider avoiding overly humanising chatbots in scenarios where self-presentation concerns are likely to be high, such as online pharmacies or stores selling personal care products.

Using non-anthropomorphised chatbots makes customers feel more at ease. By balancing clear identification and minimal human traits, companies can create a more comfortable and judgment-free shopping experience for their customers.

This approach can lead to practical benefits, such as increased consumer engagement. For example, consumers were more likely to leave their email addresses after interacting with a clearly identified, non-anthropomorphised chatbot compared to a human agent, indicating higher trust and comfort levels.





Source link

Chatbots: How they help avoid embarrassment online


Imagine shopping online for a personal item you’d rather keep private. Would you prefer talking to a human or a chatbot?

Recent researches show that chatbots — when clearly identified and not overly humanised — help consumers feel less embarrassed during such purchases. The paper “Avoiding Embarrassment Online: Response to and Inferences about Chatbots when Purchases Activate Self-Presentation Concerns” by Jianna Jin, Jesse Walker, and Rebecca Walker Reczek, published in the Journal of Consumer Psychology investigates this question.

Role of ambiguity

The rise of artificial intelligence (AI) in customer service has revolutionised how people shop online. Chatbots, a type of conversational AI, often operate behind the scenes, sometimes with ambiguous identities that can lead consumers to believe that they are interacting with a human. This ambiguity becomes particularly relevant when self-presentation concerns are at play.

Self-presentation is about controlling the impressions others form of us, especially during potentially embarrassing purchases. Jianna Jin and her colleagues explored how these concerns affect consumer interactions with chatbots versus human agents. They hypothesised that consumers with high self-presentation concerns would see an ambiguous chat agent to be human to brace for potential embarrassment.

The study confirmed this hypothesis: consumers with higher self-presentation concerns were more likely to assume an ambiguous agent was human. This inference serves as a psychological buffer, allowing them to prepare for any potential judgment, even if the agent turns out to be a bot. This aligns with the principles of Error Management Theory, which suggests that people make biased inferences under uncertainty to avoid more costly errors.

Comfort of Knowing

The research also examined consumer responses to clearly identified chatbots. Contrary to earlier findings suggesting negative reactions to known chatbots, Jin and her team found that consumers preferred clearly identified chatbots over human agents when self-presentation concerns were active. This preference stems from the perception that chatbots possess less “mind”— less capacity for consciousness and emotional judgment — compared to humans.

When consumers know they are interacting with a chatbot, they feel less embarrassed because they believe chatbots lack the emotional and cognitive depth to judge them. However, the study reveals a crucial nuance: the design of the chatbot matters. Anthropomorphised chatbots, which exhibit human-like qualities, increase consumers’ feelings of embarrassment. Thus, less human-like chatbots are better suited for sensitive purchases.

The Perfect Chatbot

The findings have significant implications for businesses aiming to improve customer experience, particularly in sensitive contexts. Clearly identifying chatbots can reduce consumer embarrassment and facilitate interactions that might otherwise be avoided.

Businesses should consider avoiding overly humanising chatbots in scenarios where self-presentation concerns are likely to be high, such as online pharmacies or stores selling personal care products.

Using non-anthropomorphised chatbots makes customers feel more at ease. By balancing clear identification and minimal human traits, companies can create a more comfortable and judgment-free shopping experience for their customers.

This approach can lead to practical benefits, such as increased consumer engagement. For example, consumers were more likely to leave their email addresses after interacting with a clearly identified, non-anthropomorphised chatbot compared to a human agent, indicating higher trust and comfort levels.





Source link