The Batch: Trust in AI, Meta's 3D Pipeline, Virtual Spaces, Baidu's Multimodal Models, and Robot Coordination

TLDR;

Public trust in AI is low, impacting adoption and development. Meanwhile, Meta unveils an open 3D generation pipeline, World Labs introduces editable virtual spaces, Baidu releases new multimodal models, and Google DeepMind enhances robot coordination through graph neural networks.

In-depth Summary

Public Trust in AI

Recent reports from Edelman and Pew Research indicate widespread distrust and lack of excitement for AI in the U.S. and parts of Europe. This public sentiment can hinder AI's progress through slow adoption, stymied projects (like data center protests), and restrictive laws. The AI community needs to address these concerns by ensuring AI benefits everyone and by avoiding hype and fear-mongering, which have damaged public trust. Link: Edelman (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0Hz5m_5PW8wM7ks6lZ3kRN4FlTWW8wmRNN5_82WnPMNCPVbqpGB9jWXC8W2s5t7f31qPhzVGk_bZ34szcQW3zHr6B4HMm0TN6QCFgNClgj9W1w7WSy77qKy9W4RNkc55vxT0WN1W-13lMT_61W3-B3gw6YjkmYW7jmnmv8kqTzYW3g0YXw4_DK5VW4B6G_L4whd40W1Q2v1C1pC08lN3xcqrcPbbkWW5Z-bVT14DGlZW1fGCDY8f_YDZW6X32KY3g2wbFW6sXqdS691sq-VDJflW7TGthtW56hhk33qWmkBW1v8KTQ8nk9sVW7jDYYk9hn5ZtW8SdmBq6j2wtdW3SmZf09h82vtW5WS-_h89Wj_dW5TP6nr5xrJ7BW2XSLsR3MyQX5W6DXZS98H4LR4W2KPHx113-NJPW4gb5Jq5_610gW8zk7Pq4rmTZtW6Sx3FG4F-j_FVNvz019gTZLWW4QXCKR7mTDGyW59yVF45jh9wGW5KBH9L79RM6MW842SBr86MsztW6q2fPP6tyT1nW2pz_b15gVq-kW8Gsfqg5Bd0kDW2gdKHC7Jk4yPW6Vxtjp1TNGDnf8-_PRs04) Link: Pew Research (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0FP5nR3bW50kH_H6lZ3nTW7mM5QK6_43rnN1z9l11qZNvvW6VZhDr32dgfhN8k1VcqFxt5kW46W8Vc2RPXp3W8GhNhM1ykx26W2PsRn92qwzd9W7D80NZ26znggW5RYB2L8ZsDT3W8CnkMk8VSl2PN80RfjZVR1PPW3X2wkd97rlq-N1Lzprg9cTcdVSV7kq6Z_NbhVsG-d-5jhpKqW3PK7DX2688WMN1K54bsvmPsZW7Jhc6N1c_vLBW8J0LtT5bZnX9W8tNj0s45srx0W18gM9r1GrfY0W6FFdDY8hbFPDVZ3Xph7lkzTZVRnjDy82TY0DW3GhKYX8cPJLvVgqG5L5ZctsJW6ylbwZ1DMbW5W523PNV7fxhqhN5b9VdmhwRz2W6TKgzQ3-Z2v0W8WvHtT8bzjXpW72v7jH2zlFrtdCYlR004) Link: local protests (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0G-5nR3bW7lCGcx6lZ3krW8Bk60_8kNlDPW2s_T3C3JJ7-fN863LYF2ncFmW4wb9mx1PM-xKW4jT5581_zWCcW3x7nF95yvhQ5W1FDH5467vvchW8kWLQ457YlZXW1zf1w43xMm6rW3f4MsY6vPqtZW152bb42fMvlbW37CWDJ3p6LHVW80NPr66_8pGDN7lxlfLsXJWLW7pXB0N7J0QKhW6VNLzC3JLxWFW8QQlQN14_-Y4N7wFN0f4pHRpW1YqjWV1mXD5fW6sQFxp4dNP-fW4r6kVW2bm3NrW6VD-lz3xG4Z2W1nkHqr4_9p6BVPy2yc9f-t23W3Lrq0v7gzFdqW1t7lhX1yDlb2W30v2mt173-vBW4C1bPG6GsnhLW69fnwX3w63TQW9bw2lP63yZv-W4ns0Fm89MR-DW7VM-_v2cm56gN7s32hlD_C0SW257L_x81LbbvW5CFqPG5LQW1zW5t_kxS6-RxpfW1vXzv07pBbGpW6pd0Y3819T3kVjVBVw3tpwnLW68hjwD8Ldyw-f9lcVgM04) Link: risk (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0FP5nR3bW50kH_H6lZ3lxW1lzRYZ383NPhW1cl0L_8wFwtyW12383c8JTfVWW8-nLms90sDfZMv-b1g8-ZsyW7PDj635wGdC3Vsq4H91tqxbWW6pzsNs3GsF9-N2Z8pcW6kp_KW2vHT6y3rnvSqW18Wdhb1R9nzpW3yfdP12n0-qDW2t6TcV65hjz0W8-0tmC7cglSbW7qcV5F8D6NQ8W1J1mTJ7TTjGlW8WKKvF8QHNgfN40TB4xHzKtRW1CNtXl1mD64QN4fdgCYtfqy8W1yRc-y44Nc9SW8bvs1F8gd1N9W1DJ4w05xF7hnW1kD8hH34X58KW3C7cN75b7YjlW301rmk6vHMdsW4BJTNh4m-fRjW6v54tT5Z0WZdW3fShH94YgtbtW1cLSvB5pvtN3W4Pq9dJ201sxlW192fnP7yqT_xf3tqSdH04) Link: article (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0Hg3qgz0W7Y8-PT6lZ3pWV_ty0g8pFZ0yW7Pr-w95fmd1PW68-NW183VgpwN6prDXfqmLQxW5PQ_Pc485bDNW1GPSS58VmtkJW7ZBkRD3B_sCLW4fPBKM673kQWVl_tgh37rsfNW98XnyN1tg02sVDL5218q9HRZW6G2nfs2mjZStW2Hj2T914JM-rW38Wxtv25jksDW6GbltZ64CrskW6zX12k5_zDWRW3_yRxq2QmHzbW1kYp4d5_zFgqM-rMRTyQtnQW3KQsnn4pSx6lW47242t2c_vB2VsJPP92FZYLKVJs8_z8YpyQRW6xflqT3wjSSPW8GJ-CX48gcYKW2QvZnm6g1RtZf9b33d404)

Open 3D Generation Pipeline (Meta)

Meta has evolved its Segment Anything Model (SAM) into an open-weights suite for generating 3D objects. SAM 3 segments images and videos based on text input, outperforming most competitors in Meta's tests. SAM 3D converts these segments into 3D objects, with human judges preferring its outputs over other models for both objects and scenes. SAM 3D Body produces 3D human figures from images. This unified pipeline simplifies 3D model creation from images, driving innovations in user experiences like Facebook Marketplace. Link: experiment (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0G-3qgz0W7lCdLW6lZ3ltW1KnwdP1qvzLlVN4J3p2F6TdcVQ02dm93ZQ4-W4431Xs42B1wWW2w8rpX4-LX5WW3YXk2p3_y-vjN5Z69kX8v2t9W6_fXMt5C5D79W8jgS5b3N6bn2W39CW0S3cZHHzVk9TNk4_8hF5W1rXm8z1xkvjwVCtDB-2dYrkPW4LRGY9240BhKW7Y2NdK2znrd0W95F13T3p7y1zW7ZNM3S14rk-PW7hM9X85NrZGZW6lsZ6T5DjXCRW6py74z87gwn4W7W822w8Th0krW12CTvW94kv8NW2JJnmB1MNJZ-W7_n9298vHlpyf3C3xvR04) Link: SAM 3 (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0GH3qgz0W6N1vHY6lZ3p5W1SHCf_7n_DShW8bYbJl658ZNHW8qJgKt5TblpmW3mgPVm1YQPz5W597F8w8sxtKWW1P6QQg6QgnCcW91v4px2CyRSdW5jxShJ7tSkSdW5DHnn12Fx--lW16hW2222QXrW2Btf9w1SD0dnW3mzK7W3ry9SJW1_fp_p24JglWW6Dl_hy6rZJnbMgHdqGjwDVqVysnQr6D-B6YW54L4d-6Q0_pyVytWvL2Xh48vW1WLXnD3XYBb_N1QwRrrJSNfKW4p09qj3HN-RpVbjkq233J5Qtf96VSbM04) Link: Weights (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0GH3qgz0W6N1vHY6lZ3mTW9h18Cf3F8VgWW6s6_MS6b6nppVwldmc7q9LzyW3YGnz43t0ZnRW2zXcJH69tyy8W71kN9g6qzkdnW6_7F6C2s3KB3W7v4LBy15-WG6W1KNcJR5m4wk6W653kK12S53CXW4xpvKd22_5QQN8cld2575HtrVndxL26FFQ5W1pZ_bK69LCwkW3yWgvt4Zc6B7N6TbxxRHf7X5W4dM5My1n-rS6W8bkCmT7X4ScfW763rvM5r0PhKW4xpM9G94YCB7V3ZR2B8nTr2kW5mfljB6b56nqf4rFrls04) Link: fine-tuning code (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0G-3qgz0W7lCdLW6lZ3kGW8FMJZg1-P8F6W8h3zST2zsGQSN1wmmVD1yPk-W26yg8s77LDFcN2_BKXdqlhtfW7R3fL68wM9kGW5_cNtV8vzpYJW8B6s1W6s1zY3W6HW7X18_BBGJN2m0pL05WWLNW4JzB3H1gCzCvW9dTkcf5N70L9W7gjgq_5NmM_SW3n3YNz2vJKRTW4rq_Cd6_YkgLW2GzPpH1SnFJRW2pX6jm2mRRBRW3hBft357QgYgW4dfgzF1y8MZQW3thdpP4VrGbDW4RDfZM6JL7bNN2339y3W_xSlW5TvfJq34GYCRW3G-kYf1db_3xf5NMzxW04) Link: license (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0G-3qgz0W7lCdLW6lZ3nFN7W_PXbM2mFGW6Y_DSt2LB5P5W9cmvRV6gnp14W8k22_61KRnXrVskWzf2rFL12W1ypj3g4tshwJW27ywfB7vg9NVN1_Pc0gJ5N6_W5Jw0xd5nVCQHW8tzc9h5g_S4rW7pKGFd58X1yfN5bL6zJJBLZWW85v4WM57TtwzVM-gPP3DnqyRW8tchsl1kznn3W74yjF36_KbVQN4cz19XhrFVBW9jgghC5KlWR8W2S6K3h8L9smbW8YM-9R16NLCNW1JKPc_4j5D4xW7VPLW18RFdFKW5gq6n63bJvtwW7VgrXC89_Tl2d6NCyP04) Link: model (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0GH3qgz0W6N1vHY6lZ3mQW7LSsPc3yH5wbW1KHBGT7RP2fgN4bbZRgqhrYvW3w_bgy2b3FlyW2gTrgF8c-7z4W5lfRw383jn1yVWKCrD8x-yQrW2fCpDv5KpR-5W41FH6N92NtLjW6wvb2h4Rsg_MW4-8Yxm7qTslZW1z0PDb7NNk39W3y90Mw6hT0jLW1sJ_sY7Ht79hW28wrPk17nGRNW71zg5W2Gqzj_W5gn2YG1fjszNW7g3Lr17CkJf8W4gQ24P6dsj-HW1K9vdW5v87ylW1KLcn13f1nmHW7V0P4y8-m5Rsf2_lRs004) Link: Weights (SAM 3D) (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0GH3qgz0W6N1vHY6lZ3nJW2TQZWz8qWsyZW1dzmg51vspR2W8gFjL95RvlCtW78PKXB7PMGvNW1bGcf92pq9s6MckVGQyQh11W68GbGB2k1QstW25lZ7H3-RHMTW7wwXbw2tWRjtN8XmtpfGqYsxW2ydJ_Q1S5zXpW5cF93R8CmPpCW5l63Rp88YjvVW6fp-0_10LHz2W81TslM75WC9rW2PJBBh5W5NrBW4f_gpX2jQVQwW7xwjBY5nCpmcW1l7kTm3g-NZyW50jJFt2L8rffW51mdY56xCcMTW4dhjwg4YXK6pf3x-NQP04) Link: inference code (SAM 3D) (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0G-3qgz0W7lCdLW6lZ3nPW5fRyFR5_P1SyW64bCLr2KBZ6lW8-2y306YmKn1W1lSZSx3pCkc4W1-tWdB3-_1vVW3t4JTf8K463JW7H8MhV46mr9_W310K6m5QkS02W3gP_Q510qYlJW5VfTfq4d_qKlW4M0zhz7J0-XrW3g2SDf7hL7y2W3Tl5Ly219kknW4RMR6W1Byq48W6MVBbr2FQbLQW4H26VW8y_dPLW3wMsnC4QrwKJVf9NjR33c1BYW7TsXqR8L-441F8F83fB27WRW1slgSK6kgyLNVT2yQj4BZcjDf3-bLcF04) Link: license (SAM 3D) (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0Hg3qgz0W7Y8-PT6lZ3l6W6cVzFM5yBhdzT9wsP9g9f9mW6D-X0b6L77byW22SFMm6Lz32XW5GRGyl7lmGNpW4BxJT33BnTWfW6jnTWY3tljTCW2Wkch06yfVzkN8fMn2472nLzW887pjM61qfDbVhGNLk5XMglWW46M2pr65c5SRW7qFK5M4FbXJLW1-wznB2-VG_QW4hP8wQ2NBlrmW85qtSQ8kZ8GSW2kbkHr8LCq2nW2hwpjd7Gw1HXVjGnsz8lgW8WW8nf_c95XthCBW7yY2-S78GMl1W85gZlK5r3M49W3b3BCQ5yRTQwW4bZSXZ1f09H8W2FNLJd8RSZLpW8zpfhr8R51xpf9jQYVK04) Link: model (SAM 3D Body) (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0FP9k468W50lwYY6lZ3llW3Q-wvf25p25HW2bLtS69k3nz6N2Vpbs1GZnMZW8CVz_L5LKpTjW5sjTs14BLHqLW1WP7wl97MKf5W38G3x-3X24BkW784n8356WWjYW4wwm0M80BV2FW4FQXcl8bCC-9W2W0jkX5x_RwSW8Y1r256QyyjCW29d4Rj669lQ3W4ZLQbt4MkMqnW6XgSyC1MvPghW87j6506J3sJKW1CRRY_3BCQZbW1ZbMP27G-6w_Vc130Z2L0KNPW2ZqRyf7Y7FtrN5YLdzD8l2FrW8ZnjyG7tWwkMW658ZBl4kfX2lW4tsDW66dHHyDW1c2QTF17YJxwW5V4fN65VCC7MW3N3gSB6PgLPRW26CQqZ6hlXW5W45T27d5qD3BJW1H2mpw4zY2PKW20v60T3STDMdVHxN1h50ZFMtVhZBC490tQRJW4d7FCv10N0xrW3C4szK1fMw2CW9gL74F2rgKvLW1cfHLL7S152jW1PpJBx1Q_51FW3xnscW7-llflMYVPpTt92CWW2d159Q5h2kVdW1Gl3G12q8PzDN8FtRHNbjjv7W6Z4VPn4z2dDJV4nMjZ7jQXxZW7HgnXv1hhczpW8p8R925Pn3RGN7hkxxxgGmdtW4018lJ8dzmN-N6jsyq5tDX_SVvzy7B5CN4cGW6WDKB64lZhtYW3P3Mcf3xM5M0MFGfHGKHxGHW2ncCS74zQDW4W9cmftf6MqydlW7X3nry4k3qntW57DrbR4RcBTRW2xxLJP1WJXhWW6qyxC52Hc0dVW2tjZpG7sp4BsW65Kvcg9hTP7dW5QKjWw8nz5TFW2Q-D468LHtc5f16KzVg04) Link: 3D objects (SAM 3D Body) (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0GH3prCCW6N1vHY6lZ3nHW50HpD429F2MjW6lr9Bf7R4j_wN92CsDXHGGRpF4wBSjR-VdNW2NrQzw7sw_6mW1Fdf8S5klwFXW8rLSn87CTl7XW4Hl0898BBmy0W1MkqCy2_cdJfW1-hNYw3csdZLW7pGFg14z8Xq6W39vwYJ9gYcydW3nX5zn7nXwLLW34hHzq4kh-jQW6qSHqX73C3w5W21YgKV2gPSpcW4_vHKl6QfWtcW3bcMd-1P5fhcW4Lw4Bp3lShHTW5ghZQR66108XW15pJN95LnK4qVZZCdW7ZKkVVf5KFFLg04) Link: EMDB (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0Gn3prCCW69sMD-6lZ3myN4xzrgF1XgKcW2Zfr2D1FnN7JW5qCSXG5FX_SKW4VKvlN8H8VYhW16PFws8TRKKtW50TRTC3_M60QW38yB2F838ClcW7Yjjq19l1Th1W4Zg19p40vrqSW9jXfW83NNgP8N16dzjVvHgCVW3Rmxf98pjRjjW5RFz_47hl19qW6pdhd046Vn9bW2LbVLr66mYY1W5N7j5K6kD1RlW5WGpXV5D8YXyW3f1VZG94wBglW5HT8KK8XPkgmN3Z3PzfXq86nf6Wjxtb04) Link: Neural Localizer Fields (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0Gn3prCCW69sMD-6lZ3lZW21qwkl30qbCjN7SNnmPDwf8zW43y4LS3tF3-_W6jl6lG1r-sPnW1z5yTR76bFbnW420qPR7wDNlfW7Hbry896RCZ8W257s3t5lKDcpM7Sc90bV7MQW17s2Ww67hX1zN4YMg7R5xDmfW5LPgXh42p11WVLfdkg1PD6kYW5fKZXh1cd-dLW5B2wLf1tBdCyW3wLpn49bsb4QN6FBNL8l-FtXW5W4jHc1fYKz7W3njM7v8hnRfvVMc2zY1ZXKpMdklbtl04) Link: Freihand (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0Gn3prCCW69sMD-6lZ3m0W1N-F0j3cR44YW9kjmt912GsTlW5FBplf22zdxxMtggTC4F7TtVyZJtg1xFRkgW16QWhf5DNf1gW6Y6gQm6Y9rQMW2fF4Wl2L9h_kW45N4l56PnjRwW7w0g3J4YR986W4xxCcP5PWCjbW1VSRHC72rQ3YV9pDnb6Zp8gfN5Wt1br_6S-2W3_ScbT2_3xbGW9k30GY6_D_wKW3t5d6-78M0dNW19XQFR7nMX7pW4VtM6l7KpH9VW8Nwc8M7Ftczkf7ywpZl04) Link: Weights (SAM 3D Body) (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0GH3prCCW6N1vHY6lZ3nKW5bdLTB42QW9JW5Fpw_z1K3th5W7zpDCx2pMlwDW96snP_5HCNkVW59r7V62-dQrkW8rnYBF8zWyrkW39Zw5r7Y6-wrW2bWjWR6j8xsrVhMgHN9bBc43W2WrYw_2l6GSMV3zXgN21KXNrV8Bp4c8Kmqy3N5-kG23fTVc8W6nwDYg62wmrpN5HDmzgZFTKzN6Nk6Cn3Pb1HW6JHLtw4_yLxHW5Rjf8D2cwzKbVJrFd48gjS87W4nj7FT8MtmQSW7gdmrC6ZMFzgN53pCwpy71Ysf2rlVqC04) Link: inference code (SAM 3D Body) (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0GH3prCCW6N1vHY6lZ3nzW1SSTQF9bCHBBW2KVPP92Z0DKcW7p-M2R2JwkvNVVH0wx7rBprwW3RR4zc85pbkGW6h_Qd97LvPSQW4yjr1g1qJChFW2J1ShQ371J_BW60tpC53fbj8-VbqmFX4GWxdZVBbW-58gSrwqW8TMGKS30vSlHW8BGHty5SMzc6N7n-lpl6gSG8W9cflmJ48qDgfN1TYMg3lPBTnW1HdbwR5btcjlW3Y01jz172PVkW8p-8Yn4xzYzhW1FlVDt19xwsVW1YDBjF3Lq3XsW6zBQTY3-pYyjdCmyCH04) Link: training data (SAM 3D Body) (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0G-3qgz0W7lCdLW6lZ3krW3X6ngy7VGcJMW25DpWX5qBJ7pW8RwCG-5qPsPkW58N8zj1gM8SDW40h73N56q031W9j-nLL8Mc29FW3ry1808Bfyf_W2GkbQf1lZ78GW2z0mPJ3zKMPNN1Td2NSsySqmV2hY_02-sHLcW2xhSQS2sCzFzW7wFgxM7wtTS9N2_fz155hFRjW61CLJz1rVwnyW6LDNFP5W59thW1khbNs24w2mbW7Fgtn12DTNz_W3XvHzf5ZHn8FW2F6Z9R6ZJ9ZgVjJsYR6wpgxKN3vgLh0NWPTmW4gs8F710rggrVLpNtd47RqDxdMB5Hd04) Link: license (SAM 3D Body) (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0GH3prCCW6N1vHY6lZ3nWW6_4YFR84dTyCW6Y7WZF2Z-08hW59n2Ys5zRZpgVQBWVl3VS633W6cp20p3G7MdNW7crPxC7bK0vNW53XTYw2qpj3YW605WNF5gZS5kW5pjPCR6Xcv5sW4m64km2d7xF1W8xlyXc4MBcv_W3zgMPh7v5dWzW3JpqDQ2f5Ll5W2r2kmL5CjqCXW6PwWHr3Wc8CKW5BKV9x13kSwMN6x_5QrJXC7JW6jzHKr7TlN1kW9jrzPk3XF2RnW5Kvdnz2nC5-9W7NWGcT6cnlD8W7V0D8Z6GMFM2W6bB0DS1BmswgW23h14c1WvqKGW2GC32b1n29q8f6lgCl404)

Generated, Editable Virtual Spaces (World Labs' Marble)

World Labs has launched Marble, a new model that generates persistent, editable, and reusable 3D virtual spaces from various inputs like text and images. Unlike previous models that created transient spaces, Marble's outputs can be exported and modified using its integrated editor, Chisel. Users can expand and combine spaces, alter styles, and download outputs in formats like Gaussian splats or meshes. This advancement aims to catalyze development in spatial AI, similar to how large language models boosted text processing. Link: Marble (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0Gn3prCCW69sMD-6lZ3pSW5CC-2q1m0VfGW1cwFlY7dpK4jW7LYSyn3B-wNKVYzgfp8D2_n8W5GM-yv1GYJJ5Vt8pmv42SwBZVFHK445CVmjyW35Nc8y6-bbbqV9LnMw3xf21jW80k_4n2v7b6kW63lKKc6t-ShCW4BXQyv2XlgxJW7hDC_k4TFxtJW7PGHgG736_t7V4fJy44DcvZfW5swyFQ4V7C-TV9nHVp6mfdMmN5q-NVl7mqwtVksmTB6gnzW_W1kPsvP4kRpkpf4cLHmq04) Link: Subscription tiers (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0Hz3prCCW8wLKSR6lZ3kzW4DVyXY8ntlkpN5XN7pkJCLCsW7x9dYk1XfFFfW4d5gnb6Xw5LqW7h_6MX2ZVfh1V2CTJx8HrbhVW56K11n8HZmMMW5_PX9C65kMYGW6zTJfJ8KJtyGVhwwBB20s9gNW1qsLpL5N-CRdW21FNBP8jhk7qW2N5RQg45XhGFW5BpMp99d_3ZDW289DdB5K297BW4GFGM391b1WXW1fFKNj7Z_W4VW2l88H_52tLrmW5sn4cy6Fmh4vW6RZLXn8mzlCbW12_twf6Kk2rpW7S7BTL79h882W4_Bg862CzVTdW6hd2gy5C1cLpW9dDr9z5yz2WVW1cdzKT7H9RqKW1PH4q097Qry5W2TXLyD5CTbBcf7P6Lg004) Link: report (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0G45m_5PW5BWr2F6lZ3ncW8GvDbt7hktSPW87W2B61vZYHFV7BMdH4xqltWW8HBRZ37h28j0W8wlk6v9d0tFrN3DhfKhrzXZMW5PYGFG33RsSwW1r_Q8f1zL_dXW9d4DWW6cGwTyW700rBB8ZCTJYW2-czqJ3GT7twW5Z9-xH18-b13VB6gYH5xS3g2W697gkp4fTXq-W41mk021qSrBbW889fwp2h8bL4N8gBJVPfnkZdVMGH5z7dmRQ6W7P8KcT7N69vyW1XK3Vx8WhFZMW2419_M8hblYqW3jvM0X8BFkdsW80XCVW2tDT_MN6HJVmMY8KPjW1z2Wzr1Dp2z-W9hH-5m6-q-_XW4X9p2G3d3D60W8nwzqw28pBqWW8Q5jql3QNTtLW7fp5vy5KFKJnN94WQdLD4GzTW8dnrZm812QG7W5hm-v12PVTjQW93dJsz6vjnytf10mmjd04) Link: said (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0G-3prCCW7lCdLW6lZ3kyW8lp-2_7LsTNPVQ0sM08YZmC0VxVBrj6gQBsSW6NYS-63JLzxcW2LykPp8Bp1ntVjrF1R4NyS_CVtXJtc3dsJjRV96qhx6QjzSnW2V4B2g7LWbP7W2Tchms1GSStZW5BCNWl56b-cKW4THLZc44vXzFW46Z8Yv5cnS62W5H-zxW5h5vV1W8vgwy_7cCdcMW6Ccc2x8BJ67cW8jPkWf1MK7b5W3XQ7Wr6btqMZW2XvwBj3hNYKnW8-D8R029YZW5f8Yxj-804) Link: RTFM (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0GH3prCCW6N1vHY6lZ3pTW9c0QRZ9dZQ1tN5npd15TRmCbW5NP_VN1SLyJHVd3QS96MN47vN5QGb4dfx4qbW6X61T93hpmvcW2xckmx4jYj3lW4nkY2h4CSVn9W4TnxfK1Hj03LW2xHRnW3pZyclW51XDpd2HrsqxN7n5N8MVDkZVM2HVQ9c3t8FW44Kqxm4Z0yKZW4lWKmj46JKNYW6pLXj36ZTmxpW7jmc7L93mKxVW2_Md6p6j_1WjW4chcwP1D6CLBW2TwM-s72LsblN2BC9jWWnw7-W4mMFZy8wHVHFf2QbCW004) Link: argues (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0Hg3prCCW7Y8-PT6lZ3mdW1KM9bm5LPLpCW4chpw95Hw_LzN7qnYbnXS2RwW4cGJZQ6mj1kQW13FbPX5GVDN5W5Qdywr6zcl0nW7vVVjd3jd4qZW79mfGx4zx1vjW2GyJRT8vH9SsV47Yrs1cSk37W6PKgPw26zvQ9W6vC6kt6V8z2NW49Prmv1lS_qWN87Qrh31KL8gVwHMQ39kqMBBW2DTsfj54vq03W1rySHz3sD3x0N4CTtXStTCdtW39QFRd1-N4R4W8DDh-b44hNw4W7HtSvT6nzWm0N598JWKbF-vVW4Fx_dS15-DwgW46c_QZ2B2sdKW4dWL0l2rVWTpW8Ky8GP30Lmt5f6JPSV204)

Baidu’s Multimodal Bids

Baidu has introduced two new models: Ernie-4.5-VL-28B-A3B-Thinking, a lightweight, open-weights vision-language model, and Ernie-5.0, a larger proprietary multimodal model. Ernie-4.5-VL-28B-A3B-Thinking excels in visual reasoning and document understanding, outperforming larger models like Gemini 2.5 Pro and GPT-5 in some tasks. Ernie-5.0 is designed as a natively multimodal model, trained on text, images, audio, and video concurrently, and reportedly performs comparably to Google Gemini 2.5 or OpenAI GPT-5. However, Ernie-5.0 has shown issues with tool calling. While Ernie-4.5-VL offers cost-effective visual reasoning and flexibility, Ernie-5.0's performance may not fully meet expectations, even with its innovative multimodal pretraining approach. Link: model (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0Hg3qgz0W7Y8-PT6lZ3pWW6ymYlC2H3JfdW3PZ3Wy8mzzhkW86xFT26pDJ4WW3Q0k233fSKFmW4yMQMZ4j6FwJW54fCPN541_NNW7s1PFs8mV38QW5CYNJQ3gYpSMW6V5klb7rvx8pW1M8pTZ8x0ZtGW36-Q8s4y3mP5W2YZQGf2XsjKqW5R5gqf83fjTjW6SfKtp5RTFS6W4m2kL96MjNc9W107Bt37Rn7mqVL_rCv995sXCVFd8Kd2ThKlrV-61Hq209RflW1hTlbb4frysKMB_fqT60t80W4NVy5v7kfZpmW993W4L4lQPjhW2Jl1vB5ZrlVkN1NFSw59mj3gW6pwHd-4WgrLGdlgP-W04) Link: mid-training (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0Hg3qgz0W7Y8-PT6lZ3mTW5bdrC87zC7WxW1SP5r42VkTN1VsyCjP2cN5z6W1zhYc81gd3VW6jfDvF86yL5RW3KgVb42753sTW7lmStt1XLwWvW84Jx9D84L6nzW3bFRnq75kJRTW1qW4v45M138dVH1rbq5SGnLFW2yd1XQ536h7vW3DLxGQ1LpNHWW8SzZMz8x3nHcW8mGXGz479TlJVDxrZt2wVGlMW55xwy72my6NRVjxmR17bFClgW5_n-rr2nLz8cW21kzGX49lPpdV41zZm2KVwRzV1Z-CR69Z-fhW2YSXc16C83lsMFZ_gYt-KmyW3vfJr51gLmByW2GR_Vy4zGzp3f1sv66n04) Link: GSPO (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0GH3qgz0W6N1vHY6lZ3m0W8RTkfv4QwCpBW5yKbGs29p8HqW6Lx-275cd6ZXW1pLqp652V61FW7Sq5604MyQmrW7TYkVq7Byx81VYHxjZ1kJT9LW1XYc6z840FmKW81tlz-65Z7mjW4L8syf8dbSLkW5n1TWw5-m8rRVv0fvQ7qHHk1W8yTnYp3Mnm77W5tqCcT6RyWDgW6YQLjh5BmkCQW5fjS6H11H5SpW8K6kYZ1HnhX4W60Rmwt942J8NW4shBrX7nPGb0W6DcPxw7mcgjYW3Zd-mp8-00kVMY4MK7-X_9GdcGJSq04) Link: IcePop (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0HT3qgz0W95jsWP6lZ3mWW5mJFvB8xB5MKW9bjPY_6hMnJMW5sz4607d-1-0W7V41W06pKJw1VMRWCV8xwbv9W5Q8Szw23ZqzmW3X7f1r8rdGdFV5rTV83vrL0WN23Sdmcdm3bBW8lNQFB4BYS_3VNsW4W5QnQJ1W68KPS895Z7RFW46RjF480fsNvW7YQhzY1j2RKBW6sb4l162d5dNW73-8TS5tM_hGW9cZM9L1PtNCMW5PqGN48JqlxYW56-FLf7Kfp1-W905Rrf24-71nW6tQNSK1--9TSW2VfcR18YzCgVW5rXCV8593RYXW4xw5xy5kMQfGVRX7XZ91200QW81pSHW5cZcjdW97g-dC8jBntGW87c4vQ3zr2XFW1bmG-z2Wrq_FW6PGb144WvBNqdkG07T04) Link: HuggingFace (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0Hg3qgz0W7Y8-PT6lZ3mCW2NlsC436yRXxW3NQt_N7M4n1MN7fG0CgDFbBMW3KvdYw1ykjdgW535nxn6nYzTvW72QrMV8LK4mGW1sX2nG4RPSxXW6XsXDV71_-GVW2sdh3z5_BhGlW7y10Vs2fxWxZN27sDfQ4JGqrW2kPZ7F8ndpSHN439rjt_6VvcW7dcj-28jP4kmW3wv7-g1sbG5wW5NySLC5MmzkqW1bcxcx1C4zHvW6Qh0J12yQnc4W1DmmDn71WwH3W944Sg790lxGqW60rzhb8HJ0mrN97ms1ff-MFqW3dwrHl2C4DY5N5k7SVhHCnJyV7VzHr93MbqTW8Lpvkr6Cndrzf40_BGb04) Link: Baidu Qianfan (Ernie-4.5-VL) (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0G-3qgz0W7lCdLW6lZ3kWW2LJf342LYRDWN3-q8lt7c9wgW7TrltX2cSpHXW4xyd408T6c2vN4k91tgPGjG2W5x1yJ-7Y_nwcW8tXFlK9kZzlyW8vsRJ72jq_xKW2Ydjcx18LVRvTHwxB63wBvKW2FrHp23VvC46Vt5wqW2VkKjxW95b7BH3MtZycW5V5WW46gmN6tW5lh99V56bRdmW6jbjm22Jp6b2VVKmmR8D8MWwW928bxK2gyWkHW57mM-57pZXw-W20SvF53Bj9nsW5TV2Nm89mRcjW3WJHd778YbRLW6Zb4yR25rnxPW62C5J_96KWN5f5-BWFd04) Link: MM-AU (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0GH3qgz0W6N1vHY6lZ3pRVgZLsW4rhtwDW2kmC6v3tB5fMW6WbPgT3QwngsW4SdDTy1yDkmNW2WcfV73_QY_8W16Lyft5FWQYmW2bLK1X5L38j_W54Nn3c4F147vW8lYWlf5FcL_CW3tg8gN89Gv4SW5mXG3x90nNnjW9jXwGs75z0s4W4wn-cm6ZQ085N47L_xC9qvccW7wL2_1785TKcN6bf85wdBg2hW7tDf8j5HtbQXW5wrT9z2kfH_tN3dQns14ZX2XW1v46yY19g0--W2sBWrh6cFKnmW2jfm7L5Jgpn-f5YsKrg04) Link: TUT2017 (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0Hz3qgz0W8wLKSR6lZ3lYN23MwJCFHGFYW4vSZhk8MHvyHW4048py86cJcFW5PxYwx2h9tKnW3GxWvh4rvr9FVKhlXs50dDRvW7xXwrS5_ZqnyW6-htQn752KmvN14nj-KDqcJpW8cnJyK9hgmXKW8vqWR81nFchNW3gyLmd2x6kFBW6z2-NR5zpHZ1McMwwXclJJfW2GwbXJ5xxjCYW7bxXD33mhkmPW46sxCb8gvTg4W1RglY-770D6WW2q_xh149dFDnW7mv7-04FWPL8W5t7Hvx7LCZWnW9kd2VZ7dnXx6W2xRVsT2wN8b7W1vw2Hy9fSl8hW73K3zc5jZF4zVX1cRT727j6FW1GqzS63VnSdsVx9yZ57k-TTtf5DdlZK04) Link: web interface (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0GH3qgz0W6N1vHY6lZ3pPW7knC103Mr1jHW55pnqD1F3jWRW4YhMQM5sT2wDW6p3Fys11B-1XW4fJtx_5FtmZCW3X4ggL7ZtRcPW1T69zR6g_gJ5W1rDSqd8g_BYqW6Yzhgm43_N1xW6hZTd69fHMvxW46P2pF1dljg4W8nBBlX1GGR83W4kNRxW1NrfM8W8hHsdw4Qx6mSW4R24xl3nWSfNW3wVjwN7T7D6kN61qFrVr5dKCW2wHTww36Vqx4W4bq6nt7BW1hfW4d1dJp40BMXSW55ZKcm7McRcYW2H1yhy91mR95f5HVWkF04) Link: Baidu Qianfan (Ernie-5.0) (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0G-3qgz0W7lCdLW6lZ3l2Vk4FdJ8Qj3r6W7wtRvV64vtR9W78LjjV1MKL2zW4-FnDb6mBtDMW4xp_Vk9k9LLlW1NJY7C5JRh3ZW5H106F7GcH9NW7hSmHL9j3VYfN4f2jr0hsG5gW1wpnQH1MWFktW3KzsSL7LS6vjW4gkZHK2c2qJHW3D6gdq3GxSyDW750_Xd4L370HN8GQGn1F-0QQW4vl2v58NRWsHW6fsv8z2SmX3tW3mdGsz93R6JTW8r27Wd6__ZZNW2s2DMX8l_TXsW38h4CY6L9MgtW9d1krl89KFLQW3419k06ZpWlDW49p4wL20rGr7f7L2Msj04) Link: reported (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0G-3qgz0W7lCdLW6lZ3phN4N9DMtQ2C2gW8PV-Zx6N8wGwW4CRGnG3_fWPYVRb0177xSqpKW3Fk_ll7Hp42hW6SPZNJ18KytgN6RC0xkSS2ymN4Rv6Zdq98tvW8RQ2Q56QxJCsN15kBS5mgF4qN94_6c2QdgTnW98K76G5tKrSGW2_GD0F53LZ85V1lLWh71Pww6W9crnwr23BVGHW4-Db9C7sXqGwW4RMV0W5Pfq_DVsC_2m8DTzfVW70WLK93v4fhZW9cQbLs8nZRNkW4b_RQp4lj3fmW98pjXq1ZxgRLW1K2YG_55-PQ-W3R1z7q7Q88Ecf4N2sLK04) Link: acknowledged (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0G-3qgz0W7lCdLW6lZ3p8W80rTRp9c-bSMW3f3rFw6nfcNGW6qpZ_T3ryPCbW54QJR062SsrLW1P3fjP2S1n7TN4QnmYqYvx-PW7Wdnrw29k962N4hLhrlR37QWW83SVWV90NWTRW7k_J2k3zMVRmW488m7K52m3yfW4Y3GwJ7X66TyW86m5dr2FRxFjW7CqW9j1rgpFdW59LPF58Rz3m7W3DCy3t8RTHfLW342CjL4VTDx-N3R3FmPD_LhbVs21Nj7Z21SWW5SQlL652mWnPW3zzQrt7tn-S-W1nxmzs1yDTMyVJpdgj3khDkrW2MHlRg3HwKgdf1GSWXF04) Link: expectations (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0Hg3qgz0W7Y8-PT6lZ3pjW7l_w822X1cCbVWVnd62VkrSgVsPv805w209xW8qPB5C2kX9PrW2PS-P_7FL1z0VM13WP2NjMFSW3tGjPd1KlQSPW5xv-fB7kl0Y_W5hPHfS1xPXmwW2-v26d1pn9CHW77gMK48fz-rqW79xVzH374h-qW2tszWs84n9znW1GXF0K7BbfqHW5dqZ-_77rKb8W1fFXC85_2lRGW6pbVKv3sl-jsW5jTXk73QyHJBW6n7Zng8clqG1W6cgmcN8dDj7fW80qFq-80tCHCW2JWrfz7wXb8mW7Jk8dB7LgNy2W71Dn4D7PS6SMW7K978s6QNH5qW1_lYC_29flmYf57d3zj04) Link: LM Arena (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0GH3qgz0W6N1vHY6lZ3psW4sw9zW8Fss8jW4_B-Wq4hN1zZW4p_fWP7h5H_9W75BZ9q97MyY0W2BBbQt1v-7svW83sRbr2kVfRJW4NbgWq3JG--QW869X1Y6VSx4VVzb97L769Qd-VYf1vf7dhrKvW5hmyT028pbbTW7vCBzl8KbFzsN87s1CH-8KnPW3ghWqc3HgCmVW6hGpqv3JYhl1N7qxzqhTPhKMW7D5Cx98lLGlbW1wVHd88MQVJPW5p3kW57m2FSBW1PPnft6sN_1dW1Bjt8473hn6DW77KFZJ5Ttx75f8c0j0j04)

Coordinating Robot Teams (RoboBallet)

Researchers at Google DeepMind, University College London, and Intrinsic developed RoboBallet, a graph neural network that automates the coordination of robotic arms in tight spaces. Traditional search-based planners struggle with the computational complexity of multiple robots and obstacles. RoboBallet overcomes this by learning to produce synchronized, collision-free motions through reinforcement learning in simulated environments. It represents robots, obstacles, and targets as nodes in a graph, generating joint velocities and continuously adapting to changes, making it more robust than hard-coded routines. This approach shows significant promise for industrial automation and potentially a deeper understanding of physics in virtual worlds. Link: RoboBallet (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0G-3qgz0W7lCdLW6lZ3l9W44ykwy1-Xz_dW7f68TR1tYyLXW6nflv83CL2H-VRFdq237jtj0W2XnyVL1nx898W6fSRlh93VzgVV7J1Q546YZlGN5s8JMXF78vxVfNMkK8PR57VW7vPsFq1HQSRgN7DMBDJB2ng0W2B9B_y1Hq-3vW2Pbm6h68ZkH6W2wNvwF7GWtlcV85G-d8ZZQ38W1rPMH06rc3pyW6lCtMN82xFTmN4rhmBP6vLpVW41_39T1zQ_6HTGs8R2tMnWnW36ydyx5B3h-gW6jWBdS69rVN4VPC2bw5nkg0vW4StMMC4wRBnvf6JNHMs04) Link: TD3 actor-critic algorithm (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0Gn3prCCW69sMD-6lZ3pnW5qP1Lm4CjCZXW57DLD06Mtk1GN1Mr2R273Y3YW8qwNP_7zBvjyW8QfTdt5K6qgBW1dC3778FKVm6Vf41hT4pGT5KW12hxNq6lDdZdW3fFJxd2bnYcWW2fRrRK97LzGMN67l_VnT3rS-N1gR9DxwVS3TW3GvS_H1v7HDvW3rN86q7PZx7mW8VBDN_3_srW5W4dMCgZ1yJxhMW158bQf3ZjqflN4SyTRPg76G-N5cmQ4WmX317W165lxk6hv-tgf80YKtY04) Link: Hindsight Experience Replay (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0Gn3prCCW69sMD-6lZ3lyW8lp-2_7LsTNPVQ0sM08YZmC0VxVBrj6gQBsSW6NYS-63JLzxcW2LykPp8Bp1ntVjrF1R4NyS_CVtXJtc3dsJjRV96qhx6QjzSnW2V4B2g7LWbP7W2Tchms1GSStZW5BCNWl56b-cKW4THLZc44vXzFW46Z8Yv5cnS62W5H-zxW5h5vV1W8vgwy_7cCdcMW6Ccc2x8BJ67cW8jPkWf1MK7b5W3XQ7Wr6btqMZW2XvwBj3hNYKnW8-D8R029YZW5f8Yxj-804)

Other News

DeepLearning.AI offers a new course, "Building Coding Agents with Tool Execution," covering agent reasoning, sandbox execution, file management, and feedback loops. Data Points provides concise AI news twice weekly, covering topics like DeepSeek’s V3.2 and Claude Opus 4.5. Job openings are available at DeepLearning.AI, AI Fund, and Landing AI. Link: Enroll today (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0Hz3prCCW8wLKSR6lZ3pyVxgm0j6mM-sQW4Pkn0M6N2qp3W3V4j-B59ystCN3c7T9QJ99tlW2p02hm1kxdKpW3slq_L1q9z1FW1Dg5kj3LgcNbW6YGSfr82NCs7W5rRjvj90TTrJW6LkQ3h2YHYQyW8fQgql67C2TfW9k8KJ811JWz7W51TTlM462x73W6KN6BM6Zht2TV66QMk80wkbZVBg_L95JtjlhW2b-T1s6C64NpW4Rq2L_1snhkkW192bV02cMXkYW89cHlJ48wzdJW9cmySd1Rz588W7HYLCv6XzTtbN3jTqf4yzJ12W80cyPP80q-_TN4K6nsGDgrsNW311q03868MMfW5mDMZh7b9sMjW5N3H2W1V6DNPf53WpQP04) Link: Subscribe today (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0G49bqx4W5BXf_W6lZ3nYM-Ptn9FZLZFN8Fb4qqD2H5vW79R2rx5ShFn6W6Kflrv2SnzRpW2VjTMT8qzRGkW2-6TL24M1d7YMN9cpRk5C_JW38Zkrq3SBn8CVp0S9l7rGkj-W7xDzzZ3HCTpBW6H79VC8BWXwrW35CbLf7LMTNJW6WnMLK5hf7ypW1qgfk51G35cmW3yvmlL7w7VCdV3fWb_1HsJzGW7FWSr41YXVJkW502SWJ1jVJqFW4_Z5R08HqKCVW30jlm68C40bXW3cd0F_3DdhgbN7B1zQYd7_XDW2nZyP97_YQBCW6f7Z_B4G7jc8W5ywhFQ55f7_WW4RwpRR5HQ7skW6jcF-W3pmwvZW2949JN6_cM20W5CDsTl4hF0SBW5G_YVp5rChmPW8f6lZX2yml5WW7p9l0S4DLV20Vx-l3n6BmLxVW3fYzqK81BVX2W6Kc1PQ4HfVyKW8HXPxQ6MyXJ6W3K6_6n8QN6JxW1pK3WQ8WwX-TV9g42C4HGZlYN3LL-Cpx3QYlN74fNjG3szBxW8-jLWR5_d4dJW3f99YQ9cXJKPW6_Ysfr7PmNY9W666SPz8_YkXpW27G7rj363YfKW65V72z4l1k0CVgdkgc3kJLsqW3mXzGD80fX46W5TStfj2p8dMzN39Dt2s7V5KMW4Dwfbs1pltZmW988KHs24qq7VW5y_2dw7x7515W32KrJM1NXDM_W5GMcwx1__0TJW4sg21Y7_01wPVfW_Y03qKqx9W3Qhx9k6pd6s_W4HNRW37dk0KzW9dXY9c5RsTfKW18L2Kc1KtJvjW23T4Xr4TzNLTN8xbnjW3Y7mJW971fJT1Mz-KfW1DdcGb1Xxyxpf4cZNK004) Link: DeepLearning.AI (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0GH3prCCW6N1vHY6lZ3nNW1Kt8G51-PlH1VXR-HJ6RqHhHN7lqyLqZZk27W9gNDYC701nGVW65kTZk7V_yZhW4Tz8fr3wT4z6W234bS52qkx3vN8xCsF_NHpxXW2JHh5w76tHRhW6wbQQB7T11K1W79Jm2r1j-0p6VTKZBz9329q4W5Xl0dP1L_0KDW2VWMk_9kq0_cW8RlKj92R92hFW4Fw-H62yCK1-W7mSZmz84pBn4VhPhqw5xRxJ8W1RW2NC5pZ4NsW2qVTzv3rjbNcW9cY0F55jb2DdW41dVLP2b0Vr1f8tqdNb04) Link: AI Fund (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0Gn3prCCW69sMD-6lZ3nzW7yd1f98xCztGW73Ssq38QwP6TW8X7T1b51Kv4qVZcTLn2-R6p2W2YJkWT58K2qYW7Zgz8j45KxV9W12nTB_95NWx9W8q6QS38MK828W6qTxd27NZ1ZxW4vvhj04CZpcFW5_T94511ZqZsV3nWly1KtvlxW8sNMWV5KM456W5SyrNz3Z4MyrW5_5hTc1gfq42W6KzTpd1LlXpXW8gT69Y1-bM7kVH5yTV2stvV5W64gXV737lkS3W1cGf5N4XhyyZd9LSQF04) Link: Landing AI (https://info.deeplearning.ai/e3t/Ctc/LX+113/cJhC404/VX2R4l4t259QW8MCB5m11CmvYW1f4B6k5GG6CmN97h0Gn3prCCW69sMD-6lZ3lRW5WFc5D5DJCwBW4YR9lX6_kLZBN8kRrSVNnQ4LN3yCVVv2XWXPW5tLLGg7lg6zFN23HwFLSwytLW5Fbwfr10xkL9W5xjnB_17jLw9N5_tyFHw2VRhVBbGhm7jQYPFW4ltglG7dsZHSN7BYbwYDVfn7W2B0-mJ4s7J5-W2tZTNP46zJ1_W2GnwCl31QX_fW7w1pQ75nDzfJW2ML5t78NtpDsW1zRlPV75w5wYW6nyFRl1lHnNTW3vD5Vk19pNpZf1kmPnM04)

Key Takeaways

• Public perception and trust are crucial for AI adoption and development, requiring the AI community to address concerns responsibly.
• Meta is advancing 3D content creation with a suite of open-weights models (SAM 3, SAM 3D, SAM 3D Body) for image segmentation and 3D object/human figure generation.
• World Labs is innovating in persistent, editable 3D virtual spaces with Marble, aiming to boost spatial AI development.
• Baidu is pushing multimodal AI with Ernie-4.5-VL-28B-A3B-Thinking (strong visual reasoning) and Ernie-5.0 (natively multimodal, but with early tool-calling issues), positioning itself against major players.
• Google DeepMind's RoboBallet uses graph neural networks and reinforcement learning to enable highly coordinated, collision-free movements in multi-robot systems, showing promise for industrial automation.

Tradeoffs and Considerations

The current public distrust in AI highlights a significant challenge for the AI community: balancing innovation with ethical considerations and transparent communication to rebuild trust. Meta's open-weights 3D models offer accessibility but require users to be mindful of licensing and trade restrictions. World Labs' editable virtual spaces are a leap forward, but the static nature of objects within them suggests further development is needed for realistic physics integration. Baidu's aggressive entry into multimodal AI with Ernie-5.0 is ambitious, though early reports of tool-calling issues indicate areas for improvement. While RoboBallet's simulation-trained robot coordination is impressive, scaling this to highly complex, dynamic real-world scenarios without extensive real-world training remains a challenge. The rapid pace of AI development, as evidenced by newer models like Gemini 3 and GPT-5.1 already surpassing Baidu's latest, implies a continuous race for relevance and performance.

Disclaimer

This article is generated by an AI assistant based on provided newsletter content and may contain inaccuracies. Always verify information from original sources.