DeepMind’s Lila Ibrahim: ‘It’s hard not to pass because of the imposter’
Lila Ibrahim is the first general manager of DeepMind, one of the world’s most famous intellectual companies. They have no experience in AI or research, which is a major part of the company, yet they oversee half of its partners, a global team of about 500 people, including engineers and scientists.
He is working on one thing, which has no meaning: to create artificial intelligence, so powerful in the human brain that it can advance science and humanity. Its job is to transform the vision into a systematic one.
“It is difficult not to have an undiagnosed condition. I’m not an AI expert and I’m here, working with some very smart people. . . It took me a while to understand everything except the first six minutes of our research meetings, ”he says. “But I realized that I was not hired to be an expert, I was hired for 30 years, my ability to understand the profession and its strengths, and I did this without fear to help us achieve this ambitious goal.”
The Lebanese-American professional, 51, joined DeepMind in 2018, and relocated his family to London from Silicon Valley, where he was a senior at Coursera’s online training company, 20 years at Intel. Before leaving Intel in 2010, he was Craig Barrett’s chief of staff at the 85,000-member staff, and had twins.
As an Arab-American in the Midwest, as well as a female engineer, Ibrahim was “always odd”. At DeepMind, too, he was a stranger: he came from work, having worked in Tokyo, Hong Kong and Shanghai. They also manage profits, Team4Tech, which recruits volunteers from technology companies to achieve education in developing countries.
DeepMind, based at King’s Cross in London, is run by Demis Hassabis and Britain’s leading team. During his three years there, Ibrahim oversaw the doubling of more than 1,000 employees in four countries, and answers some of the most difficult questions in AI: how do you make a profit and trade? How do you develop a professional pipeline in a highly competitive job market? And how do you create AI that is reliable and ethical?
Ibrahim’s first problem is how to determine the agency’s success and profitability, when it does not sell tangible assets. Purchased by Google in 2014 for $ 400m, the company lost $ 477m in 2019. Its $ 266m revenue that year came from other Alphabet companies such as Google, which pay DeepMind for any AI sales generated internally.
“Having been in the company of a number of state-owned enterprises, I know how hard the Scriptures work. In my experience, when organizations look for short periods of time, you can often get carried away. Letters should be considered short-term and long-term in terms of their importance, “says Ibrahim.” Alphabet sees DeepMind as a financial future in AI, while it also provides marketing. Get WaveNet, which is a DeepMind technology now integrated with Google products [such as Google Assistant] and entering Project Euphonia. ”This is a communication project with ALS [motor neuron disease] patients are able to keep their voice.
The software is developed primarily through the DeepMind4Google team, which works to market its AI to Google business.
He also claims that DeepMind has as much rights with its company as it “needs right now”, and makes, for example, its operational goals. “I have to tell you, when I joined you I was curious, will there be problems? And there never was, ”he says.
Another major problem is hiring researchers in the competitive market, where companies such as Apple, Amazon and Facebook are battling AI scientists. Anecdotally, it is said that the great scientists could be paid in the $ 500,000 area, while a few would rule millions. “Too Deep [pay] it’s competitive, regardless of the level and responsibility you have, but that’s not the only reason people live, ”says Ibrahim. “Here, people care about missions, and they see how the work they are doing is advancing the work [of building artificial general intelligence], not only in themselves but also as part of a larger effort. ”
The third problem that Ibrahim looked at was to define what to do with DeepMind AI research. In addition, researchers point out the dangers posed by AI, such as autonomous robots, and others such as disclosure of social discrimination and privacy attacks through technologies such as facial recognition.
Ibrahim is always driven by how technology affects social media. At Intel they worked as an internet broker for remote people in the Amazon rainforest. “After talking to Shane [Legg, DeepMind co-founder], I went home to think, can I work for this company and sleep my kids twins sleep at night and know what moms do? ”
DeepMind’s sister company Google has been criticized for meeting ethical challenges in AI. Last year, Google reportedly fired two AI researchers, Timnit Gebru and Margaret Mitchell, are said to show that language-based AI (which Google re-creates) can challenge human language. (Google reported Gebru’s departure as “resignation”Social disruption has created a problem of trust among the AI team: do technology companies like Google and DeepMind be aware of potential problems with AI, and do they aim to reduce this?
To achieve this, Ibrahim set up a multi-stakeholder group. It meets with industry research teams to discuss the risks associated with DeepMind’s operations. “You need to keep this in mind. . . and the decisions you have made and changed your mind based on this, ”he says.
He added that “if we do not have the expertise on the table, we bring in experts from outside DeepMind. We have brought in people from security, privacy, theologians, psychologists. It was a barrier to the culture of [scientists] open it and say ‘I don’t know how to use it, and I’m scared to compare, because what if I make a mistake?’ We have done much to keep these meetings mentally safe. “
DeepMind has not always been smart: in 2016, it developed the most accurate AI solution from video, and software that can be available to the deaf and blind, but it has not approved security and privacy for the public. However, Ibrahim says DeepMind is now more focused on how it can be implemented on its own devices, such as WaveNet, its voice and texting software. “We thought about the possibility of misuse. Where can we reduce it and reduce its forms,” he said.
Ibrahim says part of the job is knowing what AI cannot solve. “There are some areas that should not be used. For example, monitoring activities are a concern [and] dangerous weapons of independence. ”
He adds: “I often describe them as a good path. Everything I did prepared me for that moment, so that I could use the highest level of expertise to date, and [on] understanding. . . how to use it. ”
Three questions for Lila Ibrahim
Who is your leadership hero?
Craig Barrett. I was the chief of staff at Intel, and he was a senior at the time. He followed in the footsteps of Bob Noyse, as well as Andy Grove and Gordon Moore. . . were the myths of the semiconductor manufacturing industry. Together, we engage in the pioneer work, as we have access to the Internet in remote areas that we have never had before. He said, “If someone gives you a bad one, tell him to come and talk to me, because I have given it back to you.”
What was the first lesson you learned?
There were many people within the organization who were asking questions [my work]. I was in trouble with some of the [Barrett’s] direct reports, senior supervisors. He calmed me down and said: “Lila, the finders always have more arrows behind them than in the front, because everyone is always trying to find them.” He said, “Let me shoot an arrow so you can run it fast.” With that in mind, I want people to try not to be afraid of making mistakes. All I can do is because at the beginning of my career a military commander treated me.
If you were not CEO / leader, who would you be?
The first job I wanted was the US President, but probably the most ambassador these days. Gathering people, and understanding their differences to improve things I realized I had always wanted. It is about finding the similarities when the obvious is different.
Source link
