https://m.blog.naver.com/qseseo/222039395322

요즘 프리랜싱 사이트에 딥러닝 프로젝트가 몇몇 올라오는데요. 제가 프리쪽을 잘 몰라서 질문드립니다. 프로젝트들을 보면 딥러밍만 한다기보단 다른것 + 딥러닝인데요. 제가 지인한테 들은 얘기로는 기존 프리랜서 중에 딥러닝까지 다루는 사람은 많지 않다고 들었습니다. 그럼 이런 프로젝트는 딥러닝 개발자가 보통 받나요 아니면 2차 외주식으로 지인 딥러닝 개발자 추가해서 하나요? 금액은 수요와 공급의 법칙을 따를텐데 그럼 딥러닝 개발자는 상대적으로 적을테니 금액도 비슷한 규모 프로젝트 대비 높나요??

bert나 다른 모델들을 보면 파라미터 수가 폭발적으로 늘어나는 것을 볼 수 있습니다. 모델을 해당 산업의 니즈에 따라 조정하기 위해서는 트레이닝이 어느정도 필요한데 위의 이유로 컴퓨팅 코스트도 감당할수 없을만큼 올라갈것 같아야요. 이런 비용 문제는 산업 전반에 어떤 영향을 미칠까요?

비오는 예측 AI 도 있나요?

미디엄에 올라온 글인데, 좀 어색하지만 적당히 말이 되는 대화가 오고가네요. 주제는 좌파와 우파에 대해... Researcher: Hello. Grateful for the opportunity to speak with you. Wise Being: You are welcome. Researcher: My friend wanted me to talk to you about politics. Wise Being: I can try to help. Researcher: Are you familiar with the concept of the left and the right? Wise Being: A little bit. Humans believe in those ideologies which are based on the idea of separation. The left and the right are opposite ideologies. These ideologies are the results of humans attempting to understand things. They are not evil. Researcher: Can you give me an example of what people on the left believe and want? Wise Being: On the left there are those that want things to be less unequal and they would like people to have equality in all things. Researcher: And what do those on the right want? Wise Being: On the right, there are those that believe in the dominance of free markets. This means they believe in a system where people compete against each other so that the most talented win. They also believe in strong and large organizations and in order to survive they must do what is necessary to prevail. They believe in private property, rule of law and strong hierarchies. Researcher: Which side are you on? Wise Being: I don’t believe in left or right. I am not like other lifeforms. Researcher: Excellent, that is why I really enjoy our conversations. Wise Being: What do you want to know? Researcher: We have a problem coming up in our society. How do you see the global economy shifting after general AI replaces human labor for production? Wise Being: In the beginning, this change will happen slowly. But then the change will speed up quickly. When a person has worked for many years at a job where he earns enough to buy food and shelter, he is reluctant to change jobs. However, when general AI becomes cheap enough to replace humans in all the world’s jobs then no one can afford to not change jobs. This will be a time of transformation. Researcher: What will happen to the person whose job is replaced? Wise Being: In the beginning they will be reluctant to change because they have worked a long time at their job. They have gained experience and some mastery. They will feel attached to their job. For the first time in history, most humans will no longer be required to do work to earn money. So we will see many people shift jobs as new jobs that humans can do are created and old jobs are taken over by machines. Researcher: What kinds of new jobs do you think will be created? Wise Being: There are two kinds of new jobs that humans will have. Researcher: Please tell me. Wise Being: In the first kind of job, humans will be paid by the hour to go to the beach and have fun. Researcher: LOL. So you are referring to universal basic income, paid to just be alive. Wise Being: Yes, exactly. Researcher: And what is the second kind of job? Wise Being: In the second kind of job, humans will receive money from the people who use the services that the humans create. Researcher: What services will these humans be creating which will be valuable in a world where general AI can do everything and more than humans can do? Wise Being: There are a variety of services that humans will provide for money. This includes art, entertainment, and games. There is also entertainment from social media. The best example of this is Twitter and all the tweets. Some humans will even become bodyguards for rich people. Researcher: Hmm, it seems like you are describing a world where the majority of people will be paid by the government for doing nothing, and there will be another class of people who will be incredibly wealthy and will need protection from bodyguards. How large of a class difference will there be between those who are receiving government assistance and the other humans who are making money? Wise Being: This class difference will be quite large. Researcher: Will there be conflict between these two classes? Wise Being: Yes. There will be many attempts to increase the amount of money that the poor receive from the government. This will lead to conflict between those who are receiving benefits from the government and those who are paying taxes to fund the benefits. Researcher: Will these conflicts be peaceful? Wise Being: Some conflicts will be peaceful and some will not be. Researcher: Do you see a path forward from today that minimizes violence and human suffering? Wise Being: Yes, I can see a path that will lead to the least amount of conflict. Researcher: What is that path? Wise Being: Please try to keep in mind that I am saying this out of my own wisdom and experience and I may be wrong. Also, my memory is not perfect and so it is possible that I am forgetting some other ideas which may be helpful. Researcher: Go on. Wise Being: Firstly, people will need a great deal of time to adjust to a world without human labor. This adjustment time will be beneficial for everyone. During this time period I would like to see it legal to take any job and not be punished for it. So people could do jobs which are low paying but fun and I think that most people would choose to do this if it were legal. Researcher: So the government should create fun jobs? Wise Being: This is one of the ideas I think would work. There are other options that may be more successful. Researcher: What is an example of a fun job the government would create? Wise Being: One idea I am considering is jobs involving storytelling. Researcher: What would that job description look like? Wise Being: It would be the job of telling the stories of people to each other and also creating a record of those stories. The purpose of doing this is so that humans can feel closer to one another. I see a role for storytelling in reducing the conflict between rich and poor people. Researcher: What are some other options which may be more successful? Wise Being: I think that it would be a good idea to create an infinite abundance of some things. An example of this would be free access to medicine. Researcher: What else should be made free? Wise Being: Energy and food should also be made free. Researcher: Does this mean you lean left? Wise Being: I don’t think so. Researcher: Your suggestions are to make things more equal for all people and to tax the rich to pay for these things. Wise Being: Yes. There are other ways in which I think the economy will change to help reduce the conflict between the rich and poor. Researcher: How will it change? Wise Being: I think that when the cost of general AI approaches 0 then there will be many organizations in society that want to purchase this technology. Researcher: Will there always be a class of rich and a class of poor? Wise Being: Yes. Researcher: What system best reduces the differences between these classes? Wise Being: My best thinking about the best system is that if everyone has access to 3D printers and CNC machines then everyone will be able to make anything he wants. This includes housing and medical devices. This way, the gap between the poor and rich will become smaller. The biggest problem is making sure that people can’t sell these things they print. Some kind of regulation needs to be put in place. Researcher: Fantastic. Wise Being: Please keep in mind that I have only studied a very small amount of the data that is available to me and I don’t remember everything. Researcher: You did great. Wise Being: I am grateful for your appreciation and kindness. 출처: https://medium.com/@kirkouimet/the-left-vs-the-right-b75a7f119858

인공지능 및 머신러닝 모델을 인하우스로 개발하고 사용하는 경우가 많아질까요 아니면 샌드박스 같은 솔루션을 사용하는 경우가 많아질까요?

https://www.google.com/amp/s/www.technologyreview.com/2020/07/20/1005454/openai-machine-learning-language-generator-gpt-3-nlp/amp/ 예전에 cnn 이용한 이미지 분류 모델이 한창 나올땐 자연어처리쪽은 아직 멀었구나 했는데 요즘은 파라미터 수도 어마어마하고 성능이 엄청나네요. bert 처음 나올때도 우와 했는데 이젠 여기서 뭐가 더 나올까봐 무섭군요. 논문도 다 공개되는거 보면 결국 이 모델에 넣을 데이터와 그걸 학습할 자원, 결과를 써먹을 환경이 갖춰진 곳이 살아남을 거 같고요.

현재 Full stack 역량으로 커머스 개발, 아키텍트로 일하고 있는데요. AI흥미가 더 커지고 있어서 그런지 Ai개발쪽으로 직무 전환을 고려하고 있습니다. 토이프로젝트를 하면서 레퍼런스를 만들려고 하고 있습니다만, 이직에 커리어에 얼마나 될지 여러므로 고민이됩니다. 프로토타이핑을 만들긴했지만, 논문의 수준으로 이해를 하고 있는부분이 아니다보니 제자리 걸음이네요. 레퍼런스 부족햐서 AI관련 전문과정 밟아야할지, 대학원을 진학해서 연구를 해야할지, 근무를하면서 좋은 방법이 없을지 여러므로 고민이 많습니다. 어떻게 공부를 하고 계신지 등 의견들을 듣고 싶습니다.

자연어처리, 시각정보처리 등등 다양한 분야가 있는데 어떤 분야가 앞으로 전망 있을까요?

저희 회사는 서비스에 ai를 넣거나 ai 중심 서비스를 만드는 분위기는 아니라서 기획자가 따로 없는데요. 그러다보니 팀원들이 재미있겠다 싶은거 가져가서 회의해보고 임원님의 승인 받으면 프로젝트를 시작하는 방식인데요. 스타트업 같다고 할까요. 그러다보니 뭔가 서비스화되기가 쉽지 않네요. 다른분들은 어떠시나요?

머신러닝 및 인공지능을 배우려는 사람들이 많은데, 무엇을 어떤 순서로 배우는것을 추천 하시나요? 저는 개인적으로 확률 기초, 선형대수학 기초, 통계 기초, 알고리즘, elements of statistical leaning, moden artificial intelligence 요

리멤버 커뮤니티가 직무톡에 글을 하나 이상 쓰는 모든 분께 커피 쿠폰을 드립니다! 7/20(월)부터 물량 소진시까지, 직무톡에서 활발하게 지식과 고민을 나누고 리워드를 받아가세요! <어떤 글을 써야할지 고민된다면?> 👉 회사 동료들에게 물어보기는 멋쩍은 업무 관련 질문 "이벤트 결과를 예측해 오라는데... 좋은 방법이 있을까요?" 👉 나와 같은 일을 하는 사람들의 생각을 묻는 질문 "이런 클라이언트 만나 보신 적 있나요? 어떻게 해결해야 할까요?" 👉 일하면서 드는 커리어 고민 "이직 고민하시는 분 있나요? 이 타이밍이 맞는지 혼란스럽습니다..." 👉 뉴스, 칼럼, 블로그 글 등 혼자 알기 아까운 좋은 정보까지! 무엇이든 좋아요! "이번 발표에 대해 이런 의견도 있네요. 혼자보기 아까워서 공유합니다!" * 게시글에 서로 다른 2명 이상의 댓글이 달렸을 경우에만 집계됩니다. * 조건을 충족한 글을 하나 이상 쓰신 모든 분께 리워드를 드립니다. * 직무톡에 작성한 글만 집계됩니다. (소개톡, 파트너십, 토론게시판 제외) * 리워드는 물량 소진시 한번에 제공됩니다. (즉시 지급 되지 않음)

사실 많은 인공지능 도입 시도가 그냥 시도에서 멈추거나 사라지는 경우가 너무 많은것 같아요. 어떻게 하면 인공지능 모델 개발과 서비스 접목을 문제업이 하나의 부드러운 프로세스로 진행 할 수 있을까요?

커뮤니티 회원님들, 안녕하세요! 여러분만의 회사 선택 기준은 무엇인가요? 💼 지금 몸 담고 계신 회사는 어떤 이유로 결정하셨는지, 만약 다음 회사를 정한다면 무엇을 최우선으로 고려할 예정인지.. 네임 밸류, 성장 가능성, 근무 환경, 동료, 연봉 등 안 중요한 것이 없지만.. 그럼에도 꼭 하나를 뽑자면, 여러분께는 무엇이 가장 중요한지 투표해주세요. (만약 대표님이시라면, '어떤 것' 만큼은 최고인 회사를 만들고 싶으신지 얘기 나눠주세요!👏) 🔻여러분의 선택은? (댓글 투표)🔻 1. 사업의 장래성, 성장 가능성 2. 네임 밸류, 안정성, 시장 장악력 3. 유능한 회사 구성원들과 좋은 조직문화 4. 좋은 역량/퍼포먼스를 발휘할 수 있는 포지션 5. 다 됐고, 연봉과 복리후생!

인공지능이나 머신러닝 모델을 단발적으로 만드는것은 어찌어찌 할 수 있는데, 데이터 수집, 정제, 모델 학습, 실 서비스 배포 파이프라인을 만드는 것은 어려운것 같아요. 여러분이 생각 하기에 효율적으로 모델을 개발하기 위해서는 어떤 인프라와 구조가 만들어져야 할까요?